fernando juan pérez porras
TRANSCRIPT
UNIVERSIDAD DE CÓRDOBA
EIDA3 - Ingeniería Agraria, Alimentaria, Forestal y del Desarrollo Rural
Sostenible
TESIS DOCTORAL
MODELIZACIÓN DE LA DERIVA DE SENSORES TERMOGRÁFICOS
EMBARCADOS EN UAV PARA UN MANEJO EFICIENTE DEL RIEGO.
MODELLING THE DRIFT OF THERMOGRAPHIC SENSORS UAV FOR
EFFICIENT IRRIGATION MANAGEMENT
Directores:
D. Alfonso García-Ferrer Porras
D. Francisco Javier Mesas Carrascosa
Fernando Juan Pérez Porras
Marzo 2021
TITULO: MODELIZACIÓN DE LA DERIVA DE SENSORES TERMOGRÁFICOSEMBARCADOS EN UAV PARA UN MANEJO EFICIENTE DEL RIEGO
AUTOR: Fernando Juan Pérez Porras
© Edita: UCOPress. 2021 Campus de RabanalesCtra. Nacional IV, Km. 396 A14071 Córdoba
https://www.uco.es/ucopress/index.php/es/[email protected]
pág. 3
TÍTULO DE LA TESIS: MODELIZACIÓN DE LA DERIVA DE SENSORES TERMOGRÁFICOS EMBARCADOS EN UAV PARA UN MANEJO EFICIENTE DEL RIEGO. DOCTORANDO/A: Fernando Juan Pérez Porras
1 INFORME RAZONADO DEL/DE LOS DIRECTOR/ES DE LA TESIS
Dr. ALFONSO GARCÍA-FERRER, Catedrático del Departamento de Ingeniería Gráfica y Geomática, y FRANCISCO JAVIER MESAS CARRASCOSA, Profesor Titular del Departamento de Ingeniería Gráfica y Geomática, pertenecientes a la Universidad de Córdoba, directores de la presente tesis doctoral INFORMAN: Que la investigación desarrollada por D. Fernando Pérez Porras, bajo la dirección de los Doctores Alfonso García-Ferrer y Francisco Javier Mesas Carrascosa, ha sido desarrollada con éxito y alcanzando los objetivos inicialmente propuestos.
pág. 4
Publicaciones científicas
1. Mesas-Carrascosa, F. J., Pérez-Porras, F., Meroño de Larriva, J. E., Mena Frau, C., Agüera-Vega, F., Carvajal-Ramírez, F., ... & García-Ferrer, A. (2018). Drift correction of lightweight microbolometer thermal sensors on-board unmanned aerial vehicles. Remote Sensing, 10(4), 615; https://doi.org/10.3390/rs10040615.
Datos de 2018 (JCR): índice de impacto 4.118, índice de impacto de los últimos
5 años 4.740 y 1er cuartil en el área temática en el área temática de Remote
Sensing
2. Mesas-Carrascosa, F. J., Pérez Porras, F., Triviño-Tarradas, P., Meroño de Larriva, J. E., & García-Ferrer, A. (2019). Project-based learning applied to unmanned aerial systems and remote sensing. Remote Sensing, 11(20), 2413; https://doi.org/10.3390/rs11202413.
Datos de 2019 (JCR): índice de impacto 4.509, índice de impacto de los últimos
5 años 5.001 y 2º cuartil en el área temática en el área temática de Remote
Sensing
3. Mesas-Carrascosa, F. J., Pérez Porras, F., Triviño-Tarradas, P., García-Ferrer, A., & Meroño-Larriva, J. E. (2020). Effect of lockdown measures on atmospheric nitrogen dioxide during SARS-CoV-2 in Spain. Remote Sensing, 12(14), 2210; https://doi.org/10.3390/rs12142210.
Datos de 2019 (JCR): índice de impacto 4.509, índice de impacto de los últimos
5 años 5.001 y 2º cuartil en el área temática en el área temática de Remote
Sensing
Otras aportaciones destacables que han surgido de la presente tesis
doctoral en revistas indexadas son:
1. Mesas-Carrascosa, F. J., Verdú Santano, D., Pérez Porras, F., Meroño-Larriva, J. E., & García-Ferrer, A. (2017). The development of an open hardware and software system onboard unmanned aerial vehicles to monitor concentrated solar power plants. Sensors, 17(6), 1329.
pág. 5
Datos de 2017 (JCR): índice de impacto 2.475, índice de impacto de los últimos
5 años 3.014 y 2º cuartil en el área temática en el área temática de Instruments
& Instrumentarion.
2. Martínez-Carricondo, P., Agüera-Vega, F., Carvajal-Ramírez, F., Mesas-Carrascosa, F. J., García-Ferrer, A., & Pérez-Porras, F. J. (2018). Assessment of UAV-photogrammetric mapping accuracy based on variation of ground control points. International journal of applied earth observation and geoinformation, 72, 1-10.
Datos de 2018 (JCR): índice de impacto 4,846, índice de impacto de los últimos
5 años 5.194 y 1er cuartil en el área temática en el área temática de Remote
Sensing.
3. Agüera-Vega, F., Carvajal-Ramírez, F., Martínez-Carricondo, P., López, J. S. H., Mesas-Carrascosa, F. J., García-Ferrer, A., & Pérez-Porras, F. J. (2018). Reconstruction of extreme topography from UAV structure from motion photogrammetry. Measurement, 121, 127-138.
Datos de 2018 (JCR): índice de impacto 2.791, índice de impacto de los últimos
5 años 2.826 y 2º cuartil en el área temática en el área temática de Instruments
& Instrumentarion
Por todo ello, se autoriza la presentación de la tesis doctoral.
Córdoba, a 19 de marzo de 2021
Firma de los directores
Fdo.: Alfonso García-Ferrer Porras Fdo.: F. Javier Mesas Carrascosa
pág. 7
Agradecimientos
En primer lugar, quiero agradecer a todos los que me han
ayudado de forma desinteresada a la consecución de esta tesis
doctoral que, aunque mía, también es un poquito de todos
vosotros.
En especial quiero agradecérselo a mis directores de tesis,
Alfonso García-Ferrer por su gran apoyo y Javier Mesas, Javi
para los amigos. Amigos porque además de guiarme en la tesis
durante el proceso de doctorando, me llevo un gran amigo
para toda la vida, gracias Javi, sin ti todo esto y lo que nos
queda sería imposible.
A todos los compañeros con los que he compartido el
departamento de Ingeniería Gráfica y Geomática, Nacho, José
Emilio, Isabel, Juanjo y Mara por haberme ayudado de una
forma u otra a la consecución de los objetivos de la tesis.
A José Luis Sáiz, mi responsable en I+D en Babcock Fleet
Management, por permitirme compatibilizar la Universidad
con la empresa, generando una sinergia muy importante para
todas las partes. Sin él no hubiera conseguido evolucionar
hasta donde he llegado hoy.
pág. 8
A mi familia, en especial a mis padres Miguel y Fernanda y a
mi hermano Miguel, que desde pequeño me han empujado a
estudiar y superarme en cada momento y a levantarme cuando
me he caído.
A mis sobrinos Miguel y África, por ser parte de la felicidad de
la casa cuando todos estamos juntos y por intentar ser un
referente para ellos en todo lo que puedo.
Por último, a Carmen, mi pareja y mi compañera por su
nobleza, apoyo y ánimo incondicional no se puede medir, es
incomparable. Gracias por darme todos los días esos
empujoncitos necesarios y por tu comprensión, es imposible
quererte más.
Gracias a todos una vez más.
pág. 9
RESUMEN
El uso civil de plataformas aéreas no tripuladas ha experimentado un
notable aumento en la última década, siendo la agricultura una de las áreas
que mayor interés ha despertado. La flexibilidad que ofrecen estas
plataformas, permitiendo realizar vuelos sobre el cultivo en el preciso
momento de interés generando estudios multi-temporales con técnicas de
teledetección de muy alta resolución espacial. Estas aplicaciones están
siendo posibles por la miniaturización de sensores que hace posible
embarcarlos como carga de pago en estas plataformas. De este modo los
sensores registran información de los cultivos en distintas regiones del
espectro electromagnético, que es procesada para aplicaciones de
agricultura de precisión. En función del tipo de sensor, su uso presenta un
mayor o menor grado de madurez, beneficiando o limitando su uso. En el
caso del uso de sensores termográficos, su uso aparece más limitado a
consecuencia de la tecnología empleada si bien despierta un elevado interés
tanto para su aplicación en la detección de enfermedades o evaluación de
estrés hídrico en cultivos. Los sensores termográficos de uso civil se basan
en una tecnología de microbolómetros no refrigerados, la cual presenta
cambios continuos en la medida de temperatura. Esta inestabilidad genera
una deriva en la adquisición de los valores de temperatura que debe ser
corregida. Se presenta un método que permite calcular la deriva de
cualquier sensor termográfico en función del tiempo.
ABSTRACT
The civilian use of unmanned aerial platforms has experienced a
remarkable interest in the last decade, with agriculture being one of the
areas that has aroused most interest. The flexibility offered by these
platforms, allowing flights over the crop at the precise moment of interest,
makes it possible to carry out multi-temporal studies applying remote
sensing techniques with very high spatial resolution. These applications are
being made possible by the miniaturisation of sensors, which makes it
possible to ship them as payloads on these platforms. In this way, sensors
record crop information in different regions of the electromagnetic
spectrum, which, once processed, are used in precision agriculture
applications. Depending on the type of sensor, its use has a greater or lesser
degree of maturity, benefiting or limiting its use. In the case of
thermographic sensors, their use is more limited due to the technology
used, although they are of great interest for their application in the
pág. 10
detection of diseases or the evaluation of water stress in crops.
Thermographic sensors for civilian use are based on uncooled
microbolometer technology, which shows continuous changes in
temperature measurement. This instability generates a drift in the
acquisition of temperature values that must be corrected. A method is
presented that allows the drift of any thermographic sensor to be calculated
as a function of time.
pág. 11
Índice
2 Introducción ................................................................................................... 15
2.1 Breve repaso de la teledetección. .......................................................... 15
2.2 Primeros programas de teledetección de apoyo a la agricultura ..... 17
2.3 Dificultades del uso de Teledetección en el sector agroforestal ....... 22
2.4 Sistemas aéreos no tripulados ............................................................... 23
2.5 Teledetección aplicada a la agricultura ................................................ 26
2.5.1 Teledetección UAV .......................................................................... 26
2.6 Cargas de pago para plataformas no tripuladas. Teledetección
aplicada a agricultura de precisión con UAV ........................................... 27
2.6.1 Sensores RGB .................................................................................... 27
2.6.2 Sensores multiespectrales ............................................................... 28
2.6.3 Sensores hiperespectrales ............................................................... 30
2.6.4 Sensores radar de apertura sintética ............................................. 31
2.6.5 LiDAR ................................................................................................ 32
2.6.6 Sensores para medición de temperaturas ..................................... 33
2.7 Bibliografía ............................................................................................... 37
3 Objetivos de la tesis doctoral ...................................................................... 43
4 Capítulo 1 ........................................................................................................ 45
Drift Correction of Lightweight Microbolometer Thermal Sensors On-
Board Unmanned Aerial Vehicles ................................................................. 47
1. Introduction ................................................................................................... 48
2. Materials and Methods ................................................................................ 51
pág. 12
2.1. UAV Campaigns ....................................................................................... 51
2.2. Thermal Image Processing ........................................................................ 53
2.3. Validation .................................................................................................. 55
3. Results ............................................................................................................ 55
Validation ......................................................................................................... 62
4. Conclusions ................................................................................................... 69
References .......................................................................................................... 70
5 Capítulo 2 ........................................................................................................ 75
Project-Based Learning Applied to Unmanned Aerial Systems and
Remote Sensing ................................................................................................ 77
1. Introduction ............................................................................................... 78
2. Project-Based Learning: Characteristics and Goals .............................. 80
3. UAV-RS in Agricultural Engineering at ETSIAM (University of
Cordoba) ......................................................................................................... 82
4. Systems Design and Educational Activities .......................................... 84
5. Results ......................................................................................................... 92
6. Conclusions ................................................................................................ 96
References ...................................................................................................... 97
6 Capítulo 3 ...................................................................................................... 101
1. Introduction ................................................................................................. 104
2. Materials and Methods .............................................................................. 106
2.1. Study Area............................................................................................. 106
2.2. Remote Sensing Image Collections .................................................... 108
pág. 13
3. Results .......................................................................................................... 109
4. Discussion .................................................................................................... 117
5. Conclusions ................................................................................................. 118
References ........................................................................................................ 119
Conclusiones ................................................................................................... 127
Índice de Figuras
Ilustración 1: Arquitectura de un UAS ................................................................... 24
Ilustración 2: 3D del Compass Arrow UAS ............................................................. 25
Ilustración 3: Ortomosaico de ensayo de olivar..................................................... 28
Ilustración 4: Imagen multiespectral capturada durante un incendio donde se
aprecian diferentes contenidos de humedad ........................................................ 29
Ilustración 5: Sensorización con radar de apertura sintética en banda X desde UAS
.............................................................................................................................. 32
Ilustración 6: Nube de puntos 3D capturada por un sensor LiDAR ........................ 33
Ilustración 7: Termografía sobre viñedos .............................................................. 35
pág. 14
2 INTRODUCCIÓN
2.1 Breve repaso de la teledetección.
Fotografía literalmente significa dibujar con luz. La historia de la
teledetección comienza con la fotografía, con el primer sensor que se
desarrolló y perfeccionó, requiriendo de un método permanente de fijación
de la imagen, cuyo descubrimiento inicial fue realizado por Louis Daguerre
en 1839 (Pollack & Grushkin, 1977). Su uso se extendió rápidamente por el
mundo, lo que provocó un rápido avance tecnológico en cuanto a cámaras,
lentes y procesado. Como resultado de estos avances, el tiempo de
exposición fue recortado de los 30 minutos iniciales necesarios por el
método inicial de Daguerre con el daguerrotipo a 1/1000 s en 1875. Ejemplo
de estos avances fueron que pocos años después, en 1888, George Eastman
introdujo ya el primer rollo de película y la primera cámara fotográfica de
Kodac (Pollack & Grushkin, 1977).
El proceso inicial de revelar imágenes era muy complejo, las placas
fotográficas y las películas sólo eran sensibles a algunas regiones visibles
del espectro electromagnético, concretamente del azul y el verde del
espectro visible, por ello una luz roja de seguridad se utilizaba en un cuarto
oscuro para revelar las imágenes. Posteriormente los tintes durante el
revelado se empezaron a usar para extender el rango del revelado al rojo e
infrarrojo. Este hecho fue descubierto por Vogel of Berlin en 1873 (Neblette,
1970). Diez años después se detectó el límite de 1,3 µm de ancho de banda
en el uso de las placas fotográficas, un límite existente hoy día para el
revelado de imagen (Hudson, 1969). Así, la teoría del color fotográfica se
comenzó a desarrollar en 1868 pero fue impracticable hasta 1930, cuando el
proceso Kodachrome fue introducido para las películas(Pollack &
Grushkin, 1977). No obstante, no era un proceso fácil lo cual provocó que
no se extendiera su uso. Por lo tanto, hasta que no se desarrolló el proceso
Ektachrome no se generalizó el uso de imágenes en color, en 1950 (Pollack
& Grushkin, 1977). En este contexto, las películas pancromáticas, las cuales
sí son sensitivas a todo el espectro visible, no fueron comerciales hasta 1905
(Neblette, 1970). Las películas blanco y negro con infrarrojo no se
comercializaron hasta después de la segunda guerra mundial. Dichos
hechos bélicos, han sido probablemente los que han generado unos avances
más importantes en el sector de la teledetección.
Introducción
pág. 16
Por otro lado, y no menos importante, el uso de plataformas aéreas
para el registro de imágenes a fotointerpretar comenzaron con la necesidad
de labores de inteligencia con fines militares en momento de guerra. Por
ello, las grandes guerras han contribuido notablemente a importantes
avances tecnológicos. Las primeras fotografías fueron tomadas por globos
fijados a tierra en 1850, más adelante, en 1862, dichos globos serían usados
por el ejército de Estados Unidos para fotografiar las defensas alrededor de
Richmond, Virginia, durante la guerra de secesión (Avery, 1962). De forma
análoga pero cambiando la plataforma, Willbur Wright hizo la primera
fotografía desde una aeronave tripulada en 1909 (Colwell, 1960). La
fotografía aérea y la fotointerpretación fueron una práctica común que se
generalizó durante la primera guerra mundial. Durante la guerra, se
diseñaron cámaras instaladas en aeronaves las cuales operaban de forma
independiente al vuelo, permitiendo a los pilotos centrarse en las
maniobras de vuelo y mientras tanto las cámaras, de forma autónoma,
capturaban datos de forma constante. Estas cámaras solían estar instaladas
en la zona ventral apuntando hacia el terreno de forma cenital. Ejemplo de
esto se puede encontrar en el uso realizado por las fuerzas aéreas francesas,
llegando a procesar 10.000 imágenes diarias (Colwell, 1960). Por otro lado,
en 1918, durante la primera guerra mundial, fotointérpretes
estadounidenses detectaron e identificaron un 90% de las instalaciones
alemanas gracias a esta tecnología (Colwell, 1960). Todos estos avances en
el ámbito militar se trasladaron posteriormente a la sociedad civil, tanto en
aplicaciones comerciales como científicas alrededor de 1930. Prueba de ello
son las numerosas publicaciones científicas sobre fotointerpretación
realizadas en 1940 en los ámbitos de la ecología, arqueología, geología,
forestal, ingeniería o geografía (Colwell, 1960).
El segundo gran estímulo para la fotografía aérea y fotointerpretación
se produjo durante la segunda guerra mundial. Gracias a la información
que capturaban las cámaras aerotransportadas, se extendió su uso masivo
por las fuerzas aéreas estadounidenses llegando a capturar más de 171
millones de negativos durante la segunda guerra mundial (Infield, 1970).
En paralelo, el origen de la teledetección no fotográfica se desarrollaría
también durante este evento bélico, siendo los principales desarrollos los
relacionados con tecnología radar, sistemas termográficos y sónar.
Con todo esto, algunos investigadores a partir de ese momento
reconocieron el potencial de estos sistemas aplicados a problemas del
ámbito civil; celebrándose el primer Simposio Internacional sobre
Teledetección del Medio Ambiente en la Universidad de Michigan en 1962,
incluyendo ponencias sobre materiales relacionadas con la geofísica,
termografía o imágenes radar (Moore, 1979). Dentro de este primer grupo
Introducción
pág. 17
de investigadores que fueron pioneros en detectar las utilidades y
aplicaciones civiles por este tipo de sensores destaca Robert Colwell, de la
Universidad de California, Berkley, desarrollando soluciones para el
ámbito agrícola y forestal. Este investigador llegó a realizar 400
publicaciones sobre teledetección ambiental y junto a David Simonett, el
cual lideraría la edición de la revista Remote Sensing of Environment.
2.2 Primeros programas de teledetección de apoyo a la agricultura
En 1957, representantes de la industria química requirieron a
Agricultural Board of the National Research Council soluciones para, de una
forma más exacta, disponer de más información para detectar y evaluar la
incidencia relacionada con plagas en cultivos agrícolas y zonas forestales
que esos momentos asolaban a los Estados Unidos, generando cuantiosas
pérdidas, fijadas por el Departamento de Agricultura y Forestal de los
Estados Unidos (USDA) entorno a unos 7.000 millones de dólares anuales
aproximadamente de la época. Esto desencadenó que la Agricultural Board
of the National Research Council recomendara la formación de un comité de
Teledetección agrícola para estudiar el uso potencial de sensores
embarcados en plataformas aéreas entre los años 1957 y 1965, formado por
científicos referentes en el sector de la botánica, ingeniería, estadística,
física, silvicultura o economía (Sigafoos, 1970). Uno de las conclusiones de
este comité fue poner en órbita plataformas espaciales para de observación
de la Tierra, coincidiendo con el Año Geofísico Internacional (Acker et al.,
2014) y que el presidente de los Estados Unidos Dwight Eisenhower fue
quién lo anunció. Así, este comité consideró emplear todo tipo de sensores
que dispusieran de longitudes de onda capaces de atravesar la atmosfera y
capturar información relevante para la agricultura. Además, se
establecieron las líneas estratégicas para el desarrollo de la teledetección,
como el procesado bruto de datos, el reconocimiento de patrones,
radiometría o el desarrollo del sector aeroespacial (Macdonald, 1984).
Inicialmente, los primeros estudios se enfocaron en la vegetación de
grandes regiones para detectar brotes de estrés, controlar la propagación
de las plagas y evaluar los daños, generando procesados automatizados en
cada hito de la solución. Además, con datos proporcionados por sensores
termográficos se detectarían áreas donde la vegetación aparecía
gravemente estresada (Macdonald, 1984). Así, las aplicaciones
agroforestales fueron creciendo desde 1960 a la par del desarrollo de
nuevos sensores o mejora de los ya existentes en cuanto a resolución
temporal, espacial o espectral.
Introducción
pág. 18
Con estas plataformas espaciales se mantenía el mismo objetivo que
con las plataformas aéreas tripuladas durante la guerra: embarcar una
carga de pago que capturara datos georreferenciados sobre la Tierra. A
diferencia de éstas últimas, la monitorización de la Tierra desde el espacio
sería constante, lo que permitiría estudios multitemporales en grandes
extensiones de terreno. La observación sistemática de la Tierra desde el
espacio comenzó con el satélite TIROS-I, satélite meteorológico con una
cámara de televisión embarcada que permitía a meteorólogos distinguir
entre nubes, agua, hielo y nieve (Krueger & Fritz, 1961). Este satélite sería
posteriormente rebautizado como NOAA en 1970, cuyo nombre y familia
de satélite continua actualmente capturando datos desde el espacio.
No sería hasta el inicio del programa de observación de la Tierra
Landsat donde las técnicas de Teledetección tuvieron un gran impacto en
áreas como la agricultura, geología, minería o forestal. La plataforma
LANDSAT I, equipada con un sensor Multispectral Scanner, formalmente
llamado ERST, ofrecería datos a modo de escenas multiespectrales por
primera vez y una vídeo cámara en el espectro visible e infrarrojo. Así, el
programa Landsat sería el primero de una gran constelación de satélites
que se han ido enviando constantemente al espacio para la captura de datos
en diferentes rangos del espectro electromagnético, y de gran utilidad para
el ámbito agroforestal. Fruto del desarrollo tecnológico, Landsat 3 ya
disponía de bandas en el rango de espectro electromagnético
correspondiente al rojo, verde, infrarrojo cercano e Infrarrojo de onda larga
(Haack, 1982). Posteriormente en 1982, la puesta en servicio de Landsat 4
contaría con los sensores Multiespectral Scanner (MSS) y Thematic Mapper
™, de mucha utilidad para el sector agroforestal (Williams, Goward, &
Arvidson, 2006). Entre las bondades de las plataformas espaciales respecto
a las aeronaves tripuladas se encontraba poder contar con una cobertura
global de forma repetitiva; con una visión sinóptica y resumida uniforme
en el tiempo y abarcando grandes superficies, ofreciendo datos
multiespectrales en formato digital con un coste económico más reducido.
De este modo, inicialmente se identificaron por parte Chuck Paul en 1978,
responsable de teledetección de la Agencia Internacional para el desarrollo
(AID), ocho herramientas agroforestales a desarrollar con técnicas de
teledetección (Haack, 1982): 1) cartografía de inventarios nacionales, 2)
seguimiento de los bosques y deforestación, 3) planificación a futuro de los
usos del suelo, 4) identificación de recursos acuícolas, 5) invasión de las
zonas urbanas en tierras agrícolas, 6) planificación del transporte, 7)
utilidad de la tierra y 8) cartografía de las capacidades de los suelos. En la
Tabla 1 se muestra de forma resumida, en función de las plataformas y tipo
Introducción
pág. 19
de carga de pago embarcada, la evolución temporal del tipo de aplicaciones
desarrolladas en el sector agrícola.
Introducción
pág. 20
Año Aeronave
tripulada
Media
resolución
Alta
resolución
Ultra
resolución
Sensores
activos
Sensor
1956 Cartogr
afía en
cualquie
r rango
Cartografía
con
RADAR,
Cartografía
con SLAR
Imagen
Aérea
1972 Radar Cartografía
extensiva para
producción y
monitoreo
Cartografía
con SAR
1972
LANDSAT
1975 Detección de
cambios, usos
del suelo
LANDSAT
2
1978 Detección
regadíos,
superficie,
evaluación de
plagas
LANDSAT
3
1982-84 Detección de
sequías,
monitoreo
cambio
climático
LANDSAT
4-5
1985 Mapping
LAI
1985 SPOT
1998 Cartografía
temática
verificada
1999 Biomasa 1999
IKONOS
2000 2000
MODIS
Introducción
pág. 21
Tabla 1 Evolución temporal de las aplicaciones agrícolas de las técnicas de Teledetección
2001 Invasión
urbana en
tierras
agrícolas
Mapeado
especies
LAI
Agua
subterráne
a
Plagas
2001
QuickBird
2007 Aplicacion
es para
pequeños
agricultore
s en el
ámbito de
la
producción
,
tratamiento
s y
fertilizante
s
Mapping
LiDAR
2007
WorldView
2009 Revisión
diaria para
estudios
multi -
temporales
con 2
m2/píx
WordView2
2012 Gaofen
2014 Biofísica WorldView
3
2015 Detección
de malas
hierbas y
agricultura
de
precisión
Estructura,
biomasa.
Sentinel
Introducción
pág. 22
2.3 Dificultades del uso de Teledetección en el sector agroforestal
Desde un principio, el principal objetivo de la Teledetección ha sido
alertar de forma rápida y exacta ante la presencia de afecciones, plagas y/o
enfermedades en el ámbito agroforestal. No obstante, la Teledetección
aplicada como recurso para agricultores generó ciertas dudas en un primer
momento, especialmente por la dificultad de disponer de datos en tiempo
real. Además, la tecnología de computación en los años 60 no estaba
demasiado avanzada para procesar grandes volúmenes de datos. Como
resultado de este retraso en el acceso a los datos y o la información por parte
de los agricultores junto con no poder contar con estos en una venta
temporal de interés provocó que la Teledetección no fuera tan atractiva en
un primer momento en este campo de aplicación. La consecuencia directa
es que la Teledetección empezó a ponerse en duda como ayuda para los
sistemas expertos de toma de decisiones en el ámbito agrícola, ya que los
insumos se retrasaban mucho desde la captura de los datos. De hecho, las
primeras referencias a los sensores de Landsat 5 siempre hacían hincapié
en los mercados agrícolas o intereses por parte de grandes extensiones
agrícolas para los gobiernos, pero no para los propietarios (Jackson, 1984).
A modo de ejemplo, cultivos de ciclo corto no podrían ser gestionados
correctamente apoyados en datos o información de Teledetección decenas
de días o incluso meses después, ya que esos retrasos no permitían margen
para la toma de decisiones. En este contexto, el trabajo colaborativo entre
investigadores y usuario final permitirían definir las características técnicas
deseables que debería ofrecer un programa de observación de la Tierra de
utilidad para este sector de la agricultura, destacando: a) la disponibilidad
de los datos, siendo deseable minutos o incluso horas, b) la resolución
temporal, para que los datos fueran útiles, en el 50% de los casos ésta
debería ser menor a 5 días, sobre todo en aplicaciones de riego, c) la
resolución espacial, siendo aceptable valores de 400 m2/pixel y resolución
óptimos sobre 25 m2/pixel(Jackson, 1984), pensando principalmente en
cultivos extensivos. Atendiendo a estos condicionantes, no sería hasta 1986
con el lanzamiento de SPOT-1, con una resolución espacial en modo
pancromático igual a 10 metros y multiespectrales de 20 metros, cuando se
comenzaría a tener una aproximación real a lo planteado por Jackson,
siendo el principal problema la resolución temporal igual a 26 días,
provocando que en años posteriores se pusieran en servicio las plataformas
SPOT2- y SPOT-3 en los años 1990 y 1993 respectivamente, aumentando la
resolución temporal. Aun así y pese a todos los avances tecnológicos,
seguiría habiendo reticencia en su uso, principalmente debido a la
resolución espacial, demandándose una resolución espacial igual a 10
Introducción
pág. 23
metros con una frecuencia de paso mínima igual a 3 días, además de tener
en cuenta que el coste económico fuera asumible para una explotación
(Steven, 1993).
Para cumplir con las características que necesitaban los agricultores,
como el aumento de resolución o el aumento de la periodicidad, muchos
investigadores usaron técnicas para combinar distintos de programas de
observación que permitieran aumentar la resolución espacial o espectral de
dichos programas a partir de datos pancromáticos (Yesou, Besnus, Rolet, &
sensing, 1993) o mezclando la periodicidad de varios sensores, técnica que
todavía se mantiene en la actualidad (Skakun, Vermote, Roger, & Franch,
2017).
La mayoría de los requisitos que Jackson estableció, se cumplen
actualmente con el programa Copérnico diseñado por la Agencia Espacial
Europea (ESA). Para agricultura, el uso de Sentinel-2 se ha extendido en la
actualidad ya que dispone de unas características de periodicidad (5 días),
resolución (desde 10 hasta 60 metros) y espectral (12 bandas desde 0.43
hasta 12.51 µm) similares a las que Jackson ya estableció en 1984.
2.4 Sistemas aéreos no tripulados
Sin duda alguna los UAS han supuesto una auténtica revolución para
la teledetección. Inicialmente, como con otras tecnologías de la Geomática,
las primeras aplicaciones aparecieron inicialmente en el ámbito militar. Los
UAS son sistemas compuestos por una plataforma aérea, autopiloto,
comunicaciones, estación de control remota además de GPS e inerciales
(Ilustración 1).
Introducción
pág. 24
Ilustración 1: Arquitectura de un UAS
El primer programa militar que introdujo el desarrollo de un UAV fue
el llamado AQM-91ª Compass arrow en Estados Unidos. El Compass
Arrow (Ilustración 2) era una plataforma de 2400 kilogramos, con una carga
de pago de más de 100 kilogramos y un techo de vuelo de 20 kilómetros
que volaba 2 horas al límite de altura y con la máxima carga de pago
(Papadales, Tibbetts, Schoenung, & Meier, 1993).
Introducción
pág. 25
Ilustración 2: 3D del Compass Arrow UAS
Los UAV normalmente suelen ser clasificados por las características
que definen su rendimiento, entre ellos cabe destacar el modo de despegue.
Entre los distintos tipos de despegue se encuentran los de despegue
vertical, horizontal o por el tipo de rotor ala rotatoria, multirotor o ala fija,
entre otros Por otro lado, también podemos encontrar clasificaciones de
UAV en función del tipo de alimentación (batería, combustión, hidrógeno),
en función del tipo de misión (fotogrametría, observación, arma militar,
transporte) , por el peso (>5kg, 5-25kg, 25-200kg, 200-2000kg ó >2000kg), o
por la distancia y autonomía ( <5 horas y <100km, 5-24 horas y 100-400km,
> 24 horas y > 1500km)(Arjomandi, Agostino, Mammone, Nelson, & Zhou,
2006).
La teledetección agrícola no ha permanecido ajena al uso de UAS,
permitiendo aumentar la resolución espacial hasta pocos cm/píxel, lo que
permite abordar estudios a nivel de planta (Jurado, Ortega, Cubillas, &
Feito, 2020) o malas hierbas (Peña, Torres-Sánchez, de Castro, Kelly, &
López-Granados, 2013), permitiendo solventar dos de los principales
problemas relativos a la resolución temporal y espacial. Sin duda, además
de la plataforma UAV, dentro de un UAS el otro subsistema en orden de
importancia es la carga de pago, presentando actualmente ciertos límites
mecánicos y eléctricos comunes a cualquier plataforma no tripulada, siendo
los principales peso, tamaño, alimentación y consumo, unidad de
procesado y protección ambiental.
A nivel operacional, actualmente la legislación actual sobre aeronaves
no tripuladas aprobada por la Agencia Estatal de Seguridad Aérea a través
del Real Decreto 1036/2017 establece dos categorías de plataformas de
Introducción
pág. 26
vuelo (AESA, 2017), aeronaves inferiores a 25 Kg y aeronaves mayores de
25 Kg. Para la operación con éstas últimas, se exige certificado de tipo o
experimental, requisitos similares a los establecidos para una aeronave
tripulada. Esto tipo de requisitos tan exigentes se deben al impacto en un
accidente que pueden generar aeronaves superiores a 25 kg. Por ello, y
sorteando estos requisitos operacionales tan exigentes, la mayoría de UAS
usados en cartografía y teledetección se encuentran en el rango menor de
25 Kg, por facilidad documental, desarrollo, costes y justificación
aeronáutica. Esto implica, que el peso del conjunto aeronave,
comunicaciones, alimentación y carga de pago tienen un límite de peso
muy bajo, lo que implica la miniaturización de los subsistemas y entre ellos
la carga de pago.
2.5 Teledetección aplicada a la agricultura
La teledetección ofrece importantes ventajas en el campo de la
investigación agronómica, destacando la clasificación de cultivo (Zheng,
Myint, Thenkabail, Aggarwal, & Geoinformation, 2015) y su seguimiento y
monitorización (Mateos, González-Dugo, Testi, & Villalobos, 2013). Por
otro lado, permite la evaluación de rendimiento o producción de forma
extensiva y cuantificada (Leslie, Servina, & Miller, 2017), ofreciendo ser una
buena herramienta para ayuda a toma de decisiones (Jones & Barnes, 2000).
Además, permite capturar mínimas variaciones de los cultivos debido a
que éstos son muy vulnerables a variaciones del suelo (Sona et al., 2016),
clima u otros cambios físico-químicos (Ballesteros et al., 2018). Por último,
la aplicación de técnicas de teledetección permite aumentar y
homogeneizar rendimientos de los cultivos (Sona et al., 2016) de forma que
se aumenta la producción al tiempo que se reducen costes para grandes
superficies (Khanal, Fulton, Shearer, & Agriculture, 2017; Potić, Bugarski,
& Matić-Varenica, 2017).
2.5.1 Teledetección UAV
En cuanto a las principales funcionalidades que permite teledetección
desde UAV se encuentran i) estimación de la superficie cultivo,
clasificación del mismo, fenotipo o producción a nivel de planta (Yang et
al., 2017), ii) control de la humedad del suelo, detección de estrés en
cultivos, detección de enfermedades o plagas y la evaluación de la
fertilidad del suelo (Bellvert, Zarco-Tejada, Girona, & Fereres, 2014) y por
último, iii) control de malas hierbas, vigilancia de inundaciones, control de
la erosión y monitorización de la cubierta vegetal (Peña et al., 2013).
Introducción
pág. 27
Por otro lado, este tipo de teledetección con UAV volando a baja
altura, máximo 120 metros sobre el terreno según la legislación actual, junto
con el avance de los sensores actuales, ha permitido desarrollar una
tecnología muy útil para la agricultura de precisión. La agricultura de
precisión es una técnica de gestión de parcelas agrícolas basada en la
variabilidad existente en las mismas y el manejo localizado de los inputs a
partir de un listado de herramientas tecnológicas que permite monitorizar
una parcela a nivel de píxel o parcela para mejorar la calidad y producción
agrícola (Pierce & Nowak, 1999). La variabilidad espacial y temporal dentro
de una misma parcela obliga a disponer de una resolución tal, que permita
dar respuesta a esta variabilidad. Si no se dispone de esta resolución, se
corre el riesgo de no dar una respuesta heterogénea de gestión de la parcela
para generar una homogeneización en el crecimiento, producción, calidad
o respuesta del cultivo.
2.6 Cargas de pago para plataformas no tripuladas. Teledetección
aplicada a agricultura de precisión con UAV
2.6.1 Sensores RGB
Los sensores con respuesta en el espectro visible embarcadas en las
plataformas aéreas no tripuladas suelen ofrecer la mayor resolución
espacial, pudiendo ser implementados mediante dos tecnologías, Charge –
coupled device (CCD) o Coplementary Metal Oxide Semiconductor
(CMOS), pudiendo disponer unas u otras del método de captura rolling o
global shutter. En los últimos tiempos los sensores CMOS han mejorado
mucho las características respecto a los de tipo CCD en ruido y sensibilidad,
siendo mejores en términos de bajo consumo, bajos niveles de alimentación
y bajo coste.
En general, para cartografía o teledetección desde un UAV lo ideal es
trabajar con sensores global shutter CMOS, que no presenta un coste
demasiado elevado y ofrecen un resultado óptimo con un menor consumo
(Bigas, Cabruja, Forest, & Salvi, 2006). Por otro lado, los sensores CMOS
pueden utilizar focales con monturas fijas tipo C o micro 4/3, lo que permite
ir adaptando la focal de forma remota en función de la resolución objetivo
en cada momento. Esto puede ayudar en gran medida a mantener la
resolución durante el vuelo, es decir, si el autopiloto del UAV no varía su
posición respecto al terreno en tiempo real se puede mantener variando la
focal en tiempo real de la cámara.
Introducción
pág. 28
Dentro de la categoría de sensores pasivos, este tipo de sensores son
los que disponen de una mejor calidad geométrica en comparación con los
distintos sensores del rango del espectro electromagnético, por lo que es
aconsejable a partir de ellos calcular cualquier producto que sea
independiente del espectro, como modelos de elevaciones o superficies
(Moravec et al., 2017).
Este tipo de sensores se usan para generar ortomosaico (Ilustración 3:
Ortomosaico de ensayo de olivar), modelos digitales del elevaciones o de
superficies (Remondino et al., 2011), delimitación de cultivos(Moravec et
al., 2017), mapeado de recursos naturales (Carfagna & Gallego, 2005;
Verbyla, 1995), fotointerpretación (Carfagna & Gallego, 2005), detección de
caminos y carreteras (Mokhtarzade & Zoej, 2007), cubicación de
movimiento de tierras o erosión del terreno(ULVİ̇ & Geosciences, 2018),
delimitación a nivel de planta o detección de malas hierbas (Torres-
Sánchez, López-Granados, De Castro, & Peña-Barragán, 2013).
Ilustración 3: Ortomosaico de ensayo de olivar
2.6.2 Sensores multiespectrales
Este tipo suelen disponer de sensores CMOS con distintos filtros
calibrados para registrar la reflectancia espectral en un rango del espectro
electromagnético concreto, realizando su elección en función de la
aplicación a desarrollar. De hecho, no es extraño encontrar sensores CMOS
para las que se ha diseñado una rueda con distintos filtros que van girando
y capturando datos en el rango del espectro en cada momento que le
interese al usuario con un reducido coste económico (Morales et al., 2020).
Introducción
pág. 29
La mayoría de estos sensores suelen disponer de filtros para
aplicaciones agrícolas que cubren la región del espectro electromagnético
entre 0,4 y 0,9 µm (Ilustración 4: ). Además, es posible emplear sensores que
recogen datos en otros rangos del espectro infrarrojo un poco más alto,
denominados Short Wave Infrared (SWIR), los cuales registran datos desde
0,9 hasta 3 µm (Saari et al., 2011), de gran utilidad en agricultura para
aplicaciones relacionadas con estrés hídrico (Ghulam et al., 2008) o
contenido de nitrógeno (Herrmann, Karnieli, Bonfil, Cohen, & Alchanatis,
2010).
Ilustración 4: Imagen multiespectral capturada durante un incendio donde se aprecian diferentes contenidos de humedad
Por lo general, los sensores multiespectrales se utilizan para detectar
en la vegetación variables como la vigorosidad de la planta, clorofila, índice
de área foliar, daños por plagas o enfermedades, detección de estrés,
determinación del tipo de suelo y pigmentos fotosintéticos, o cuantificación
de la fertirrigación (Wójtowicz, Wójtowicz, Piekarczyk, & Science, 2016).
Respecto a los sensores SWIR, las principales aplicaciones se centran en la
predicción del estrés del cultivo y el contenido de nitrógeno en plantas,
perjudicial para el ser humano (Camino et al., 2018).
Introducción
pág. 30
2.6.3 Sensores hiperespectrales
Se clasifican normalmente en función del tipo de captura, que puede
ser de barrido de línea o de área (Adão et al., 2017). Por un lado, las cámaras
de barrido van realizando escaneos línea a línea o punto a punto, para
luego generar en la fase de post-procesado un cubo hiperpespectral
georreferenciado. Por otro lado, las cámaras de área realizan las capturas
como su propio nombre indica, capturando todos los píxeles en un mismo
momento, pudiendo disponer de un sensor rolling o global shutter.
Este tipo de sensores, particularmente las de barrido, necesitan de una
unidad inercial muy precisa y correctamente calibrada que permita
posteriormente, cuando se genera el cubo hiperespectral, la unión correcta
de todas las líneas de barrido para formar el cubo hiperespectral (Adão et
al., 2017). Así, es de gran importancia, sobre todo para agricultura de
precisión, que estos sensores dispongan de unidades inerciales de gran
precisión que permita la correcta generación de ortomosaicos
hiperespectrales correctamente georreferenciados.
En general, en esta tipología de sensores, las de tipo barrido suelen
disponer de un número de bandas espectrales mayor que las de tipo área.
Para capturar un gran número de bandas espectrales se hace necesario las
captura por barrido por una limitación estrictamente tecnológica, las
comunicaciones del sensor hacia la unidad de procesamiento interna. Este
tipo de comunicaciones comúnmente se realizan por Gigabit Ethernet
(GigE), lo cual limita el ancho de banda a 1 Gbps. Esta limitación en las
comunicaciones provoca que cuando se desean almacenar las tramas de la
sensorización, no se puedan almacenar imágenes muy pesadas junto con
muchas bandas espectrales, ya que los datos capturados pasarían de 1 Gbps
de peso y no podrían llegar a la unidad de procesamiento. Actualmente, a
las comunicaciones internas en las cámaras se les está incorporando fibra,
sobre todo en visión artificial, lo cual permitirá en el futuro aumentar el
ancho de banda del flujo de datos, lo que permitirá aumentar el número de
bandas espectrales capturadas en las cámaras de área (Li, 2016).
En cuanto a las aplicaciones agrarias a partir de la captura de datos
hiperespectrales se encuentran el cálculo de nitrógeno en planta (Näsi et al.,
2018), contenido de carotenos en viñedo (Jeziorska, 2019; Zarco-Tejada et
al., 2013), determinación de las características hidrológica de la superficie
del cultivo (Jeziorska, 2019), predicción de riegos y estrés hídrico (Albornoz
& Giraldo, 2017; Zarco-Tejada, González-Dugo, & Berni, 2012), detección o
estudio de enfermedades como Verticillium en Olivo(Zarco-Tejada et al.,
Introducción
pág. 31
2012) o determinación de clorofila y pigmentos fotosintéticos (Adão et al.,
2017; Calderón, Navas-Cortés, & Zarco-Tejada, 2015).
2.6.4 Sensores radar de apertura sintética
Los radares de apertura sintética son sensores activos capaces de
penetrar sobre la vegetación en función de la frecuencia del espectro en la
que se dispongan los filtros, centrados en las frecuencias microondas. Este
tipo de radares disponen de grandes antenas que emiten pulsos, calculando
el retardo de los ecos, lo que les permite determinar la distancia una vez
reflejado el pulso sobre la superficie. Al ser sensores activos, no necesitan
fuentes de iluminación, por lo que pueden capturar datos tanto de día como
de noche.
La principal característica de estos sensores es la alta resolución de la
que se dispone en la dirección del movimiento del sensor mediante la
síntesis de una antena de grandes dimensiones a partir de una pequeña,
esta característica es la que impone su nombre a estos sensores.
En una imagen SAR lo que se aprecia son intensidades que dependen
del tipo de reflectividad del objeto, en función de si la superficie es rugosa
o plana. La reflectividad son los datos que se almacenan en las
polarizaciones, ya que, como cualquier onda microondas, disponen de una
polarización. En general, los mejores sensores son los que disponen de
mayores polarizaciones disponibles. Lo ideal siempre es disponer de cuatro
polarizaciones Horizontal-Horizontal, Horizontal-Vertical, Vertical-
Vertical, Vertical-Horizontal.
La mayoría de sensores SAR, se centran en la banda X que va desde 8
a 12 Ghz, C de 4 a 8 Ghz o L, de 1 a 2 Ghz. En general, las frecuencias de
mayor poder de penetración son las de menor longitud de onda, como la
banda P centrada estas entre 450 y 900 Mhz (Schmullius & Evans, 1997),
muy utilizada para funcionalidades forestales debido a que estas masas
tienen una gran fracción de cabida cubierta (Ilustración 5).
Introducción
pág. 32
Ilustración 6: Sensorización con radar de apertura sintética en banda X desde UAS
La mayoría de las aplicaciones agrícolas de estos sensores se centran
en estudios sobre la humedad del suelo o tipo de suelo(Lyalin, Biryuk,
Sheremet, Tsvetkov, & Prikhodko, 2018). Esto es debido a su poder de
penetración a través de la vegetación y debido a la reflectividad medida
que depende del tipo de suelo. Este tipo de características de los radares de
apertura sintética permiten obtener datos hidrológico e hidromorfológicos
de los suelos o determinar distintos tipos de suelos o su nivel de
humedad(Lyalin et al., 2018).
2.6.5 LiDAR
Los sensores LiDAR (Light Detection And Ranging) son sistemas que
están compuestos por una unidad láser, una unidad inercial, un sistema
GNSS y una unidad de procesamiento. El sistema láser suele ser un sistema
con dos espejos que envía pulsos para los que se mide su tiempo de retorno,
lo que permite calcular distancia. Esta distancia unida a los datos GNSS e
inerciales, permite generar una estructura de puntos en 3D (Ilustración 7:),
llamada nube de puntos (Shan & Toth, 2018).
Introducción
pág. 33
Ilustración 7: Nube de puntos 3D capturada por un sensor LiDAR
La penetración del sensor LiDAR depende de la frecuencia de emisión
de los pulsos del láser, variando entre 100 y 1000 Khz, es decir, desde cien
mil hasta un millón de medidas por segundo. La penetración también
depende de los ecos del sensor por cada pulso. En el ámbito agrícola con
sensores sencillos que permitan configuraciones hasta 100 Khz y 2 ó 3 (Lin
& Habib, 2021) ecos por pulso es suficiente para la caracterización de
cultivos mientras que, para aplicaciones forestales, es necesario configurar
los sensores desde 500 Khz para fracciones de cabida cubierta altas con
distintos estratos y hasta 5 ó 6 ecos por pulso (Jakubowski, Guo, & Kelly,
2013). La mayoría de los datos LiDAR se utilizan para el generación de
modelos digitales del superficiies y elevaciones, permitiendo además
calcular la estructura vertical de la vegetación (Zimble et al., 2003). Además,
permiten realizar clasificaciones de usos del suelo (Mesas-Carrascosa,
Castillejo-González, de la Orden, & Porras, 2012), detección para la
digitalización de infraestructuras como caminos (Buján, Guerra-
Hernández, González-Ferreiro, & Miranda, 2021), ríos (Bowen &
Waltermire, 2002) o cuencas hidrológicas(Barber, Shortridge, & Science,
2005). Por otro lado, permiten calcular biomasa, estructura de la vegetación
o cálculos precisos de erosión del terreno o seguimiento de cultivos
mecanizados (Lin & Habib, 2021; Zhou et al., 2020). Por lo general, las
mejores aplicaciones agrícolas se obtienen añadiendo a los datos LiDAR
textura capturada con sensores multiespectrales, termográficos o
hiperespectrales (Bradbury et al., 2005).
2.6.6 Sensores para medición de temperaturas
Existen dos categorías, radiométricas y no radiométricas, operando en
distintos rangos del espectro, Medium Wave Infrared (MWIR), de 3,5 a 5
µm y Long Wave Infrarred, de 7,5 a 13,5 µm. Las temperaturas que miden
Introducción
pág. 34
estos sensores se obtienen a través de la radiación de onda que emiten los
objetos en el espectro infrarrojo. Esta radiación es emitida por los cuerpos
debido a que se encuentran a una temperatura superior al cero absoluto, 0
grados kelvin (Speakman & Ward, 1998). Estos sensores, bien de MWIR o
bien LWIR, a diferencia de los de tipo multiespectral poseen, para realizar
las mediciones de temperatura, un bolómetro, el cual tiene la posibilidad
de estar refrigerado por criogenización. Las cámaras en el rango MWIR son
las que disponen de este tipo de bolómetro refrigerado, por lo que
presentan un mayor peso que las de tipo LWIR, que suelen disponer de
bolómetro no refrigerado.
El funcionamiento de un bolómetro se basa en tres fenómenos físicos
i) la radiación del entorno, ii) la transferencia de calor dentro de las partes
sólidas de esta y iii) la conservación de las corrientes eléctricas (Thomas,
Crompton, & Koppenhoefer, 2015). El funcionamiento del bolómetro
principalmente se modela mediante el acoplamiento de la transferencia de
calor y la corriente eléctrica. Los bolómetros que funcionan con
temperaturas criogénicas y por tanto constantes, pueden ser más sensibles
y con mediciones más fiables que los que no lo son (J. Thomas & AltaSim
Technologies, 2015).
El uso potencial de la teledetección termográfica en agricultura incluye
la supervisión y programación del riego, detección de enfermedades en
plantas que causen estrés, estimación del rendimiento de la fruta,
evaluación de la madurez de la fruta o detección de magulladuras y golpes
en frutas y verduras (Ishimwe, Abutaleb, & Ahmed, 2014).
Los sensores MWIR y LWIR se diferencian a su vez entre sensores
térmicos y termográficos. Los primeros son sensores que sólo disponen de
bolómetro y que traducen la diferencia de temperaturas irradiada por los
objetos en el infrarrojo a una imagen con una paleta de colores. Esta paleta
de colores se asigna en función de la radiación o temperatura emitida por
cada objeto.
A diferencia de los sensores térmicos, los de tipo termográfico
disponen de un módulo calibrado radiométricamente, similar a una unidad
de computación, que convierte los datos capturados por el bolómetro a
valores radiométricos. Posteriormente, para convertir los datos
radiométricos a reflectancia, es necesario realizar una conversión en
función de la calibración del bolómetro con un cuerpo negro (Grimberg,
2012). Por tanto, a partir de una cámara termográfica se puede obtener, por
un lado, los mismos resultados que una térmica si se integra la salida
analógica y, por otro lado, a partir de la salida digital se pueden conseguir
datos radiométricos que permiten al usuario disponer de la temperatura
Introducción
pág. 35
por píxel (Ilustración 8). Estos píxeles una vez georreferenciados tras un
vuelo de fotogrametría, permitirían una comparación multitemporal a
diferencia de una cámara térmica.
Ilustración 8: Termografía sobre viñedos
Los sensores con bolómetro no refrigerado son más económicos y
pequeños (Jensen, McKee, & Chen, 2014) que los refrigerados debido sobre
todo a la implantación de una tecnología más sencilla, lo que ha
desembocado en un uso más extendido de los mismos en aplicaciones de
agricultura de precisión con UAV. En general este tipo de sensores con
bolómetro no refrigerado son bastante precisos, con variaciones entre 30 y
50 mK, pero bastante inexactos, errando las medidas varios grados Celsius
(Minkina & Dudzik, 2009). Este tipo de bolómetros al estar en contacto con
el viento y el aire a diferentes temperaturas, van cambiando el valor de sus
mediciones constantemente, ya que la electrónica está programada por sí
sola a mantener una temperatura constante. Este principio se conoce como
Non-Uniformity Correction (NUC) (Ibarra-Castanedo & Maldague, 2013)
eliminando la necesidad de calibración constante al actualizar los
coeficientes de corrección en función de los niveles de radiación de la
escena (Olbrycht, Więcek, & De Mey, 2012), de este modo, se puede aplicar
una compensación continua a las mediciones. Este principio provoca
oscilaciones en la medición de temperatura que junto con la deriva térmica
Introducción
pág. 36
provocada por el viento en contacto con el bolómetro que recibe el sensor
mientras está embarcado en el UAV, genera mediciones inexactas. En
conclusión, estos sensores miden muy bien valores de temperatura en
relativo, pero no valores de temperatura absolutos, necesitando
correcciones para compensar las derivas (Kelly et al., 2019). Así, para poder
extender el uso de estas cámaras en agricultura de precisión mediante
UAV, se hace indispensable realizar correcciones atmosféricas que
permitan que las medidas realizadas por el sensor sean precisas y a la vez
exactas. Este tipo de correcciones permitirán calcular índices de vegetación
a partir de valores de temperatura, generalmente índices de estrés hídrico.
Esta aplicación directa de la termografía sobre el estrés de los cultivos
y por tanto en el manejo eficiente del riego se debe a que, un indicador del
estrés hídrico en los cultivos se basa en el cierre de los conductos
estomáticos de la hoja. El cierre estomático inducido por el estrés hídrico
reduce la tasa de transpiración, reduciendo así el enfriamiento por
evaporación y aumenta la temperatura de las hojas, hecho que puede ser
detectado por cámaras termográficas embarcadas en un UAV, lo que
permite cuantificar el nivel de estrés y sus mitigaciones a través del riego
(Berni, Zarco-Tejada, Sepulcre-Cantó, Fereres, & Villalobos, 2009).
Después de todas las funcionalidades citadas anteriormente, se
aprecia que las tecnologías UAV están revolucionando la agricultura,
apareciendo constantemente nuevos desarrollos y aplicaciones que
permiten la toma de decisiones en días en lugar de semanas o meses,
prometiendo una importante reducción de costes y un aumento de
rendimiento. Estas decisiones permiten aplicar de forma eficaz los insumos
agrícolas, apoyando los pilares de la agricultura de precisión. Estos pilares
se centran en prácticas agrícolas adecuadas, lugar adecuado, momento
oportuno y cantidades adecuadas y controladas.
Sin embargo, la proliferación de los UAV ha sido muy alta pero no así
la explotación real de los vehículos no tripulados en agricultura inteligente,
debido sobre todo a los retos a los que se enfrenta la selección y despliegue
de las tecnologías, como por ejemplo los métodos de adquisición de datos
o procesamiento de imágenes, que no siguen un flujo estandarizado en un
área relativamente nueva. Por ello, en el ámbito de los UAV, sensores,
termografía, etc. se hace necesario el establecimiento de unos flujos de
trabajo que permitan en cualquier desarrollo ofrecer unos resultados
óptimos y una aplicabilidad directa en el ámbito agrícola.
pág. 37
2.7 Bibliografía
1. Acker, J., Williams, R., Chiu, L., Ardanuy, P., Miller, S., Schueler, C., . . . Manore,
M. (2014). Remote sensing from satellites.
2. Adão, T., Hruška, J., Pádua, L., Bessa, J., Peres, E., Morais, R., & Sousa, J. J. J. R. S.
(2017). Hyperspectral imaging: A review on UAV-based sensors, data processing
and applications for agriculture and forestry. 9(11), 1110.
3. AESA. (2017). Real Decreto 1036/2017, de 15 de Diciembre, por el que se regula la
utilización civil de las aeronaves pilotadas por control remoto.
4. Albornoz, C., & Giraldo, L. F. (2017). Trajectory design for efficient crop irrigation with
a UAV. Paper presented at the 2017 IEEE 3rd Colombian Conference on Automatic
Control (CCAC).
5. Arjomandi, M., Agostino, S., Mammone, M., Nelson, M., & Zhou, T. J. R. f. M. E. c.,
University of Adelaide, Adelaide, Australia. (2006). Classification of unmanned
aerial vehicles.
6. Avery, T. E. (1962). Interpretation of aerial photographs. Retrieved from
7. Ballesteros, R., Ortega, J. F., Hernandez, D., Del Campo, A., Moreno, M. A. J. I. j. o.
a. e. o., & geoinformation. (2018). Combined use of agro-climatic and very high-
resolution remote sensing information for crop monitoring. 72, 66-75.
8. Barber, C. P., Shortridge, A. J. C., & Science, G. I. (2005). Lidar elevation data for
surface hydrologic modeling: Resolution and representation issues. 32(4), 401-410.
9. Bellvert, J., Zarco-Tejada, P. J., Girona, J., & Fereres, E. J. P. a. (2014). Mapping crop
water stress index in a ‘Pinot-noir’vineyard: comparing ground measurements
with thermal remote sensing imagery from an unmanned aerial vehicle. 15(4), 361-
376.
10. Berni, J., Zarco-Tejada, P., Sepulcre-Cantó, G., Fereres, E., & Villalobos, F. J. R. S. o.
E. (2009). Mapping canopy conductance and CWSI in olive orchards using high
resolution thermal remote sensing imagery. 113(11), 2380-2388.
11. Bigas, M., Cabruja, E., Forest, J., & Salvi, J. J. M. j. (2006). Review of CMOS image
sensors. 37(5), 433-451.
12. Bowen, Z. H., & Waltermire, R. G. J. J. J. o. t. A. W. R. A. (2002). Evaluation of light
detection and ranging (lidar) for measuring river corridor topography 1. 38(1), 33-
41.
13. Bradbury, R. B., Hill, R. A., Mason, D. C., Hinsley, S. A., Wilson, J. D., Balzter, H., .
. . Bellamy, P. E. J. I. (2005). Modelling relationships between birds and vegetation
structure using airborne LiDAR data: a review with case studies from agricultural
and woodland environments. 147(3), 443-452.
14. Buján, S., Guerra-Hernández, J., González-Ferreiro, E., & Miranda, D. J. R. S. (2021).
Forest Road Detection Using LiDAR Data and Hybrid Classification. 13(3), 393.
15. Calderón, R., Navas-Cortés, J. A., & Zarco-Tejada, P. J. J. R. S. (2015). Early
detection and quantification of Verticillium wilt in olive using hyperspectral and
thermal imagery over large areas. 7(5), 5584-5610.
pág. 38
16. Camino, C., González-Dugo, V., Hernández, P., Sillero, J., Zarco‐Tejada, P. J. J. I. j.
o. a. e. o., & geoinformation. (2018). Improved nitrogen retrievals with airborne-
derived fluorescence and plant traits quantified from VNIR-SWIR hyperspectral
imagery in the context of precision agriculture. 70, 105-117.
17. Carfagna, E., & Gallego, F. J. J. I. s. r. (2005). Using remote sensing for agricultural
statistics. 73(3), 389-404.
18. Colwell, R. J. A. S. o. P., Falls Church, Virginia. (1960). 1960, Manual of
Photographic Interpretation.
19. Ghulam, A., Li, Z.-L., Qin, Q., Yimit, H., Wang, J. J. A., & Meteorology, F. (2008).
Estimating crop water stress with ETM+ NIR and SWIR data. 148(11), 1679-1695.
20. Grimberg, E. (2012). Radiometry using an uncooled microbolometer detector. In:
Google Patents.
21. Haack, B. N. J. W. D. (1982). Landsat: A tool for development. 10(10), 899-909.
22. Herrmann, I., Karnieli, A., Bonfil, D., Cohen, Y., & Alchanatis, V. J. I. J. o. R. S.
(2010). SWIR-based spectral indices for assessing nitrogen content in potato fields.
31(19), 5127-5143.
23. Hudson, R. D. (1969). Infrared system engineering (Vol. 1): Wiley-Interscience New
York.
24. Ibarra-Castanedo, C., & Maldague, X. P. (2013). Infrared thermography. In
Handbook of technical diagnostics (pp. 175-220): Springer.
25. Infield, G. B. (1970). Unarmed and Unafraid, p. 6. New York: Macmillan.
26. Ishimwe, R., Abutaleb, K., & Ahmed, F. J. A. i. r. S. (2014). Applications of thermal
imaging in agriculture—A review. 3(03), 128.
27. J. Thomas, J. S. C. a. K. C. K., & AltaSim Technologies, C., OH. (2015). Multiphysics
Analysis of Infra Red Bolometer. Proceedings of the 2015 COMSOL Conference.
28. Jackson, R. D. (1984). Remote sensing of vegetation characteristics for farm management.
Paper presented at the Remote Sensing: Critical Review of Technology.
29. Jakubowski, M. K., Guo, Q., & Kelly, M. J. R. S. o. E. (2013). Tradeoffs between lidar
pulse density and forest measurement accuracy. 130, 245-253.
30. Jensen, A. M., McKee, M., & Chen, Y. (2014). Procedures for processing thermal images
using low-cost microbolometer cameras for small unmanned aerial systems. Paper
presented at the 2014 IEEE Geoscience and Remote Sensing Symposium.
31. Jeziorska, J. J. R. S. (2019). UAS for wetland mapping and hydrological modeling.
11(17), 1997.
32. Jones, D., & Barnes, E. J. A. S. (2000). Fuzzy composite programming to combine
remote sensing and crop models for decision support in precision crop
management. 65(3), 137-158.
33. Jurado, J. M., Ortega, L., Cubillas, J. J., & Feito, F. J. R. S. (2020). Multispectral
mapping on 3D models and multi-temporal monitoring for individual
characterization of olive trees. 12(7), 1106.
34. Kelly, J., Kljun, N., Olsson, P.-O., Mihai, L., Liljeblad, B., Weslien, P., . . . Eklundh,
L. J. R. S. (2019). Challenges and best practices for deriving temperature data from
an uncalibrated UAV thermal infrared camera. 11(5), 567.
pág. 39
35. Khanal, S., Fulton, J., Shearer, S. J. C., & Agriculture, E. i. (2017). An overview of
current and potential applications of thermal remote sensing in precision
agriculture. 139, 22-32.
36. Krueger, A. F., & Fritz, S. J. T. (1961). Cellular cloud patterns revealed by TIROS I.
13(1), 1-7.
37. Leslie, C. R., Servina, L. O., & Miller, H. M. (2017). Landsat and Agriculture: Case
Studies on the Uses and Benefits of Landsat Imagery in Agricultural Monitoring and
Production: US Department of the Interior, US Geological Survey.
38. Li, J. J. O. (2016). A highly reliable and super-speed optical fiber transmission for
hyper-spectral SCMOS camera. 127(3), 1532-1545.
39. Lin, Y.-C., & Habib, A. J. R. S. o. E. (2021). Quality control and crop characterization
framework for multi-temporal UAV LiDAR data over mechanized agricultural
fields. 256, 112299.
40. Lyalin, K. S., Biryuk, A. A., Sheremet, A. Y., Tsvetkov, V. K., & Prikhodko, D. V.
(2018). UAV synthetic aperture radar system for control of vegetation and soil moisture.
Paper presented at the 2018 IEEE Conference of Russian Young Researchers in
Electrical and Electronic Engineering (EIConRus).
41. Macdonald, R. B. (1984). A summary of the history of the development of
automated remote sensing for agricultural applications. Transactions on Geoscience
Remote Sensing(6), 473-482.
42. Mateos, L., González-Dugo, M., Testi, L., & Villalobos, F. J. A. W. M. (2013).
Monitoring evapotranspiration of irrigated crops using crop coefficients derived
from time series of satellite images. I. Method validation. 125, 81-91.
43. Mesas-Carrascosa, F. J., Castillejo-González, I. L., de la Orden, M. S., & Porras, A.
G.-F. (2012). Combining LiDAR intensity with aerial camera data to discriminate
agricultural land uses. Computers
44. electronics in agriculture, 84, 36-46.
45. Minkina, W., & Dudzik, S. (2009). Infrared thermography: errors and uncertainties:
John Wiley & Sons.
46. Mokhtarzade, M., & Zoej, M. V. (2007). Road detection from high-resolution
satellite images using artificial neural networks. International journal of applied earth
observation geoinformation, 9(1), 32-40.
47. Moore, G. K. J. H. S. B. (1979). What is a picture worth? A history of remote
sensing/Quelle est la valeur d'une image? Un tour d'horizon de télédétection. 24(4),
477-485.
48. Morales, A., Guerra, R., Horstrand, P., Diaz, M., Jimenez, A., Melian, J., . . . Lopez,
J. F. J. S. (2020). A Multispectral Camera Development: From the Prototype
Assembly until Its Use in a UAV System. 20(21), 6129.
49. Moravec, D., Komárek, J., Kumhálová, J., Kroulík, M., Prošek, J., & Klápště, P. J. A.
R. (2017). Digital elevation models as predictors of yield: comparison of an UAV
and other elevation data sources. 15(1), 249-255.
50. Näsi, R., Viljanen, N., Kaivosoja, J., Alhonoja, K., Hakala, T., Markelin, L., &
Honkavaara, E. J. R. S. (2018). Estimating biomass and nitrogen amount of barley
pág. 40
and grass using UAV and aircraft based spectral and photogrammetric 3D features.
10(7), 1082.
51. Neblette, C. B. (1970). Fundamentals of Photography: Van Nostrand Reinhold.
52. Olbrycht, R., Więcek, B., & De Mey, G. J. A. O. (2012). Thermal drift compensation
method for microbolometer thermal cameras. 51(11), 1788-1794.
53. Papadales, B. S., Tibbetts, T. A., Schoenung, S. M., & Meier, W. R. (1993). Remote
sensing with high-altitude unmanned aerial vehicles. Paper presented at the Airborne
Reconnaissance XVII.
54. Peña, J. M., Torres-Sánchez, J., de Castro, A. I., Kelly, M., & López-Granados, F. J.
P. o. (2013). Weed mapping in early-season maize fields using object-based analysis
of unmanned aerial vehicle (UAV) images. 8(10), e77151.
55. Pierce, F. J., & Nowak, P. J. A. i. a. (1999). Aspects of precision agriculture. 67, 1-85.
56. Pollack, P., & Grushkin, P. (1977). The picture history of photography: from the earliest
beginnings to the present day: Abrams New York.
57. Potić, I., Bugarski, M., & Matić-Varenica, J. (2017). Soil moisture determination using
remote sensing data for the property protection and increase of agriculture production.
Paper presented at the Worldbank conference on land and poverty”, The World
Bank, Washington DC.
58. Remondino, F., Barazzetti, L., Nex, F., Scaioni, M., Sarazzi, D. J. I. a. o. t. p., remote
sensing, & sciences, s. i. (2011). UAV photogrammetry for mapping and 3d
modeling–current status and future perspectives. 38(1), C22.
59. Saari, H., Pellikka, I., Pesonen, L., Tuominen, S., Heikkilä, J., Holmlund, C., . . .
Antila, T. (2011). Unmanned Aerial Vehicle (UAV) operated spectral camera system for
forest and agriculture applications. Paper presented at the Remote Sensing for
Agriculture, Ecosystems, and Hydrology XIII.
60. Schmullius, C., & Evans, D. J. I. J. o. R. S. (1997). Review article Synthetic aperture
radar (SAR) frequency and polarization requirements for applications in ecology,
geology, hydrology, and oceanography: A tabular status quo after SIR-C/X-SAR.
18(13), 2713-2722.
61. Shan, J., & Toth, C. K. (2018). Topographic laser ranging and scanning: principles and
processing: CRC press.
62. Sigafoos, R. S. (1970). Remote Sensing. With Special Reference to Agriculture and
Forestry. National Research Council Committee on Remote Sensing for
Agricultural Purposes. National Academy of Sciences, Washington, DC, 1970. xvi,
424 pp.+ plates. $12.95. NAS Publication No. 1723. In: American Association for the
Advancement of Science.
63. Skakun, S., Vermote, E., Roger, J.-C., & Franch, B. J. A. g. (2017). Combined use of
Landsat-8 and Sentinel-2A images for winter crop mapping and winter wheat yield
assessment at regional scale. 3(2), 163.
64. Sona, G., Passoni, D., Pinto, L., Pagliari, D., Masseroni, D., Ortuani, B., & Facchi, A.
(2016). UAV multispectral survey to map soil and crop for precision farming applications.
Paper presented at the Remote Sensing and Spatial Information Sciences Congress:
pág. 41
International Archives of the Photogrammetry Remote Sensing and Spatial
Information Sciences Congress: 19 July.
65. Speakman, J. R., & Ward, S. J. Z.-J.-. (1998). Infrared thermography: principles and
applications. 101, 224-232.
66. Steven, M. (1993). Satellite remote sensing for agricultural management:
Opportunities and logistic constraints. Journal of photogrammetry Remote Sensing,
48(4), 29-34.
67. Thomas, J., Crompton, J., & Koppenhoefer, K. (2015). Multiphysics Analysis of Infra
Red Bolometer. Paper presented at the Proceedings of the 2015 COMSOL Conference
in Boston, Boston, MA, USA.
68. Torres-Sánchez, J., López-Granados, F., De Castro, A. I., & Peña-Barragán, J. M. J.
P. o. (2013). Configuration and specifications of an unmanned aerial vehicle (UAV)
for early site specific weed management. 8(3), e58210.
69. ULVİ̇, A. J. I. J. o. E., & Geosciences. (2018). Analysis of the utility of the unmanned
aerial vehicle (Uav) in volume calculation by using photogrammetric techniques.
3(2), 43-49.
70. Verbyla, D. L. (1995). Satellite remote sensing of natural resources (Vol. 4): CRC Press.
71. Williams, D. L., Goward, S., & Arvidson, T. (2006). Landsat. Photogrammetric
Engineering Remote Sensing, 72(10), 1171-1178.
72. Wójtowicz, M., Wójtowicz, A., Piekarczyk, J. J. C. i. B., & Science, C. (2016).
Application of remote sensing methods in agriculture. 11(1), 31-50.
73. Yang, G., Liu, J., Zhao, C., Li, Z., Huang, Y., Yu, H., . . . Zhang, X. J. F. i. p. s. (2017).
Unmanned aerial vehicle remote sensing for field-based crop phenotyping: current
status and perspectives. 8, 1111.
74. Yesou, H., Besnus, Y., Rolet, J. J. I. j. o. p., & sensing, r. (1993). Extraction of spectral
information from Landsat TM data and merger with SPOT panchromatic
imagery—a contribution to the study of geological structures. 48(5), 23-36.
75. Zarco-Tejada, P. J., González-Dugo, V., & Berni, J. A. J. R. s. o. e. (2012).
Fluorescence, temperature and narrow-band indices acquired from a UAV
platform for water stress detection using a micro-hyperspectral imager and a
thermal camera. 117, 322-337.
76. Zarco-Tejada, P. J., Guillén-Climent, M. L., Hernández-Clemente, R., Catalina, A.,
González, M., Martín, P. J. A., & meteorology, f. (2013). Estimating leaf carotenoid
content in vineyards using high resolution hyperspectral imagery acquired from
an unmanned aerial vehicle (UAV). 171, 281-294.
77. Zheng, B., Myint, S. W., Thenkabail, P. S., Aggarwal, R. M. J. I. J. o. A. E. O., &
Geoinformation. (2015). A support vector machine to identify irrigated crop types
using time-series Landsat NDVI data. 34, 103-112.
78. Zhou, L., Gu, X., Cheng, S., Yang, G., Shu, M., & Sun, Q. J. A. (2020). Analysis of
plant height changes of lodged maize using UAV-LiDAR data. 10(5), 146.
79. Zimble, D. A., Evans, D. L., Carlson, G. C., Parker, R. C., Grado, S. C., &
Gerard, P. D. J. R. s. o. E. (2003). Characterizing vertical forest structure
using small-footprint airborne LiDAR. 87(2-3), 171-182.
3 OBJETIVOS DE LA TESIS DOCTORAL
El principal objetivo de esta Tesis Doctoral es explorar y analizar el uso
de sensores embarcados en plataformas espaciales y aéreas no tripuladas
aplicando técnicas de Teledetección en distintos campos de actuación,
desde la modelización del comportamiento del sensor durante el registro
de datos hasta la generación de información de utilidad para la toma de
decisiones, pasando por la formación y capacitación de profesionales en el
manejo de estas tecnologías. Concretamente esta Tesis queda estructurada
Capítulo 1. Drift correction of lightweight microbolometer thermal
sensors on-board unmanned aerial vehicles
En este capítulo se ha desarrollado una metodología que permita, a
partir de la extracción de puntos de control en imágenes termográficas
registradas por un sensor embarcado en una plataforma aérea no tripulada
en la zonas de solape longitudinal y transversal durante la fase de
aerotriangulación en el flujo de trabajo fotogramétrico, evaluar la deriva
térmica del bolómetro producida en el sensor en función del tiempo. Una
vez calculada esta deriva, se han desarrollado unos modelos que dan
respuesta a esta deriva de la temperatura en función del tiempo, generando
una nuevas imágenes termográficas corregidas de este efecto.
Capítulo 2. Project-based learning applied to unmanned aerial systems
and remote sensing
El desarrollo de la tecnología de los vehículos aéreos no tripulados y
la miniaturización de los sensores han cambiado la forma de utilizar la
teledetección, popularizando esta disciplina en la agricultura de precisión.
Además de la transferencia de estas tecnologías al sector productivo, no
menos importante es su incorporación en la capacitación de ingenieros. En
este capítulo se presenta como se ha introducido el uso de estas tecnologías
en la formación de los alumnos de Master en Ingeniería Agronómica a
través de la formación basada en proyectos.
Capítulo 3. Effect of lockdown measures on atmospheric nitrogen
dioxide during SARS-CoV-2 in Spain
Debido a la pandemia producida por el virus SARS-CoV-2, la
concentración de gases de efecto invernadero en la atmósfera ha
disminuido notablemente, sobre todo durante el periodo de confinamiento
de la población. En este capítulo, a través del uso de sensores embarcados
Objetivos
pág. 44
en la plataforma Sentinel-5P del programa de observación de la Tierra
Copernicus de la Unión Europea se ha analizado esta reducción y su
relación con la densidad de población en España.
4 CAPÍTULO 1
Publicado en Remote Sensing. 2018, 10(4), 615;
https://doi.org/10.3390/rs11202413
Capítulo 1
pág. 47
Article
Drift Correction of Lightweight
Microbolometer Thermal Sensors
On-Board Unmanned Aerial
Vehicles
Francisco-Javier Mesas-Carrascosa 1,*, Fernando Pérez Porras 1, Jose
Emilio Meroño de Larriva 1, Carlos Mena Frau 2, Francisco Agüera-Vega 3, Fernando Carvajal-Ramírez 3, Patricio Martínez-Carricondo 3 and
Alfonso García-Ferrer 1
1 Department of Graphic Engineering and Geomatics, University of Cordoba,
Campus de Rabanales, 14071 Córdoba, Spain; [email protected] (F.P.P.);
[email protected] (J.E.M.d.L.); [email protected] (A.G.-F.) 2 Departamento de Gestión Forestal Ambiental, University of Talca, 3460000,
Talca, Chile; [email protected] 3 Department of Engineering, University of Almeria, La Cañada, 04120 Almería,
Spain; [email protected] (F.A.-V.); [email protected] (F.C.-R.); [email protected]
(P.M.-C.)
* Correspondence: [email protected]; Tel.: +34-957-218-537
Received: 23 March 2018; Accepted: 12 April 2018; Published: date
1. Abstract: The development of lightweight sensors compatible with
mini unmanned aerial vehicles (UAVs) has expanded the agronomical
applications of remote sensing. Of particular interest in this paper are
thermal sensors based on lightweight microbolometer technology. These
are mainly used to assess crop water stress with thermal images where an
accuracy greater than 1 °C is necessary. However, these sensors lack
precise temperature control, resulting in thermal drift during image
acquisition that requires correction. Currently, there are several strategies
to manage thermal drift effect. However, these strategies reduce useful
flight time over crops due to the additional in-flight calibration
operations. This study presents a drift correction methodology for
microbolometer sensors based on redundant information from multiple
overlapping images. An empirical study was performed in an orchard of
high-density hedgerow olive trees with flights at different times of the
day. Six mathematical drift correction models were developed and
assessed to explain and correct drift effect on thermal images. Using the
proposed methodology, the resulting thermally corrected orthomosaics
Capítulo 1
pág. 48
yielded a rate of error lower than 1° C compared to those where no drift
correction was applied.
Keywords: UAV; uncooled thermal sensor; precision agriculture; thermal
orthomosaic
1. Introduction
World agriculture faces three major challenges that represent an
apparent contradiction: to feed a growing population, to contribute to the
reduction of rural poverty, and to manage the natural resource base [1,2].
Precision agriculture is believed to be an efficient method of crop
production because it is accurate, inputs are optimized leading to reduced
costs and environmental impact, and because it provides an audit trail that
consumers and legislation require [3]. Precision agriculture emphasizes
spatial-temporal data analysis and management jointly rather than
singularly [4], as well as requiring a detailed description of canopy status
and its variation in the field during the growth cycle [1].
Remote sensing methods have been demonstrated to be very useful in
monitoring large areas while remaining cost effective [5]. Traditional
remote sensing techniques have used manned aerial or satellite platforms
to measure canopy reflectance in the electromagnetic spectrum range from
400 to 2500 nm [6]. These platforms have a temporal and spatial resolution
that limits their utility in agriculture assessments due to the dynamic
changes in crops in relation to the environment [7,8]. In recent years,
unmanned aerial systems (UASs) have been used in a broad range of
applications, including precision agriculture projects, principally because
unmanned aerial vehicles (UAVs) have become more reliable, their
performance (flight time, range) has improved, and the sensors have
miniaturized [9]. Therefore, UAS technology allows for the possibility of
acquiring information with high spatial and temporal resolution. In
precision agriculture where analyses are performed of yield variation over
a field and across years, half of the variation comes from year to year
variation [10]. A well-timed sequence of UAV flights can contribute to the
analyses of spatial and temporal variations.
A variety of sensors can be used as payload on board UAVs, ranging
from Light Detection and Ranging system (LiDAR) [11,12], Red-Green-Blue
(RGB) [13,14], multispectral [15,16], and hyperspectral [17,18] to thermal
[19,20]. Infrared thermography allows users to monitor plant water status
Capítulo 1
pág. 49
for detecting stress or for applying deficit irrigation techniques [21], to
monitor for infections [22], water stress detection [20,23] or for phenotyping
plants [24,25].
To date, two different thermal systems are available: cooled and
uncooled. The first, being very sensitive and highly accurate, are used on
board satellite and aerial platforms. However, they are large, heavy, and
power consuming. Uncooled thermal sensors are used as UAV payloads
because they are smaller, lighter, and consume less power than cooled
thermal sensors [26]. However, they are not as sensitive nor as accurate.
Microbolometers are uncooled infrared radiation thermal sensors
distributed in an array. Low values of noise-equivalent temperature
difference (NETD) in uncooled thermal sensors, reaching 20 mK, have
allowed their use in applications where only cooled thermal sensors were
once suitable. However, temperature drift continues to be a disadvantage
causing unwanted detector gain and offset non-uniformity in registered
temperature data.
To combat this, a non-uniformity correction (NUC) is applied to
remove noise using digital signal processing techniques on the detector
output signal. It requires knowledge of coefficient corrections for every
detector in the array [27]. For each detector in the array, a determined gain
and offset is stored in the sensor. However, the change of the offset
coefficient has to be updated due to the thermal drift effect. This
temperature drift is principally caused by the detector’s casing, which
overheats and dissipates power and heat onto the detectors and electronic
circuits. Hence, it is necessary to perform thermal drift correction and
periodically update the correction values for each detector [28]. Without
this correction, the temperature error would increase by approximately 0.7
°C per minute [29].
The most commonly used approach to compensate for thermal drift is
shutter based [28,30]. Others have used a contact sensor between the
detector matrix and the lens [31] or other locations inside the sensor [32], as
well as “blind” pixels whose signal does not depend on the radiation of the
observed scene [33]. Other types of NUCs are scene-based methods, which
are divided into two categories: statistical [34] and registration-based
methods [35].
Prior to flight, uncooled thermal sensors need to be stabilized after
being switched on [36]. During this period, the absolute temperature
progressively shifts until it is stabilized. As such, it is necessary to consider
how environmental variables affect the registered temperature values,
especially during a long acquisition period [37], and take into account wind
Capítulo 1
pág. 50
effect, varying cloud cover, position, and orientation of the sensor during
the UAV flight. To remove thermal drift effects and simultaneously apply
radiometric correction, some authors program UAV flights to cover ground
targets with known temperatures. At the cost of useful UAV flight time,
radiometric calibration coefficients are calculated so that the UAV can
repeatedly fly over the nearest ground target and thus eliminate thermal
drift [38,39].
The goal of this study was to determine a methodology using high
spatial resolution thermal imagery acquired from a UAV while removing
temperature drift independently of NUC applied by the sensor without any
extra UAV flight operations. The specific objectives of this work were
aimed at (i) modeling temperature drift effect; and the assessment of (ii)
different drift models at (iii) different hours of the day.
Capítulo 1
pág. 51
2. Materials and Methods
The presented study was carried out in Córdoba, Spain (north latitude
37°56′05′′, west longitude 4°42′59′′, WGS 84) in June 2016, using a 10 ha
orchard of high-density hedgerow olive trees (Olea europaea L. cv
Arbequina). Figure 1 shows the development of olive trees during the
growing season. Being a typical Mediterranean region, the climate is
characterized by warm, dry summers and cool, wet winters with an
average annual rainfall equal to 180 mm.
Figure 1. Development of olive trees during the growing season.
2.1. UAV Campaigns
The unmanned aerial vehicle (UAV) used was an MD4-1000 multi-
rotor drone (Microdrones GmbH, Siegen, Germany). This UAV is a
quadcopter with an entire carbon design. The system has a maximum
payload equal to 1.2 kg. It uses 4 × 250 W gearless brushless motors and
reaches a cruising speed of 15.0 m/s. The UAV was equipped with a Gobi
640-GiGe thermal sensor (Xenics nv, Leueven, Belgium), which is an
uncooled long-wave infrared (LWIR) thermal sensor delivering raw digital
images at 16 bits of sensor calibrated radiance with a dynamic range from
−20 °C to 120 °C and a spectral resolution of 0.05 °C. It has a focal length
equal to 18 mm and operates in a spectral band range from 8 μm to 14 μm.
Registered images have a dimension equal to 640 × 480 pixels and a pixel
pitch of 17 μm. Moreover, it has an onboard image processing system to
perform a non-uniformity correction, an auto offset, and an auto gain.
However, the continuous changing conditions in which the sensor operates
cause temperature values to degrade throughout the UAV flight although
the image processing system is operating. In this manuscript, the NUC has
been deactivated, obtaining a set of images without any compensation. A
stabilization procedure on the thermal sensor was conducted before each
UAV flight as described in Berni et al. [36]. The thermal sensor was pre-
heated for twenty minutes on field before each UAV flight to stabilize its
Capítulo 1
pág. 52
internal temperature. The sensor was connected to a stick PC Asus QM1
(Asustek Computer Inc., Taiwan, China) to store the images via an Ethernet
port. Sensor weight totaled 710 g with flight duration equal to 20 min.
UAV flights were performed on 1 June 2016 at 120 m above ground
level with a ground sample distance (GSD) equal to 11.3 cm (Figure 2). Side
and forward lap settings were 80% and 70%, respectively, with 632 images
registered. Five aluminum disks were placed on the plot as ground control
points (GCPs), one in each corner and the other in the center of the study
area. Because of the low emissivity of aluminum GCPs, they were visible
in thermal images. Each GCP was measured with the stop-and-go
technique through relative positioning by means of the NTRIP protocol
(The Radio Technical Commission for Maritime Services, RTCM, for
Networked Transfer via Internet Protocol) using two GNSS (global
navigation satellite system) receivers. One of the receivers was a reference
station for the GNSS Red Andaluza de Posicionamiento (RAP) network
from the Institute for Statistics and Cartography of Andalusia, Spain, and
the other, a Leica GS15 GNSS (Leica Geosystems AG, Heerbrugg,
Switzerland), functioned as the rover receiver.
The UAV flew over the crop at 8:30, 12:30, 16:00, and 18:30 local time
in clear skies. A Davis Vantage Pro2 weather station (Davis Instrument
Corp., Hayward, CA, USA) was used to monitor climate conditions during
the UAV flights. This weather station was equipped with a three-cup
anemometer, an air temperature and humidity sensor, and a barometer.
Table 1 shows air temperature, percentage of relative humidity, mean wind
speed, and atmospheric pressure during each UAV flight on this date.
These UAV flights were used to acquire thermal images of soil and crop
under different atmospheric and temperature conditions.
Table 1. Atmospheric conditions for individual unmanned
aerial vehicle (UAV) flights on 1 June 2016.
UAV Flight (Local Time)
8:30 12:30 16:00 18:30
Air temperature (°C) 21 30 36 37
Relative humidity (%) 48 26 18 17
Mean wind speed (m/s) 2 3 6 6
Atmospheric pressure (hPa) 1022.6 1021.2 1018.4 1017.4
Capítulo 1
pág. 53
Figure 2. Flight planning and distribution of ground control points.
2.2. Thermal Image Processing
Thermal images were processed in three stages: thermal drift
correction, geometric correction, and radiometric correction. Figure 3
describes the drift correction applied to the thermal images. As previously
discussed, thermal drift is when the same location in the terrain presents
different temperature values in different images. Thermal drift correction
is based on points having a constant temperature during the UAV flight. In
the first stage, a set of distinctive features are extracted from the UAV
images using algorithms based on “structure from motion” (SfM)
techniques described by Lowe [40]. SfM techniques extract individual
features in each thermal image that are matched to their corresponding
feature in the other images from the same UAV flight. As Figure 3 shows,
a point appears in several images that belongs to different laps, each having
a different temperature value due to drift effect. Every thermal image has
a temporal reference, allowing the drift effect to be evaluated as it occurred
in flight. In the proposed methodology, sensor drift is modeled as a
function of time where each point of each image has a timestamp obtained
by a GNSS sensor from the UAV autopilot. This methodology is applied to
the set of characteristics extracted by the SfM algorithms. As a result, a
mathematical model that describes thermal drift as it occurs for the
duration of the UAV flight is achieved. Subsequently, this model is applied
to all thermal images to obtain a new thermal image where temperature
values are uniform on all corresponding points along the UAV flight. Six
Capítulo 1
pág. 54
different drift correction models (DCMs) were developed to describe
thermal drift for the duration of the UAV flight: exponential, exponential,
lineal and polynomial order two, three, and four. Each DCM was applied
to the UAV thermal images to generate a new collection of images, which
were orthorectified and then processed into thermal orthomosaics.
Figure 3. Graphical description of thermal drift correction model (DN:
digital number, t: time).
To obtain a single thermal orthomosaic of the area of interest, images
have to be aerotriangulated, rectified, and finally mosaicked. Based on
previous research results, Pix4dmapper by Pix4D SA was selected
(Lausanne, Switzerland) and described by Mesas-Carrascosa [15] to do this
processing. Afterward, digital numbers (DNs) of the thermal orthomosaics
were converted to temperature values using the information obtained by
the radiometric calibration of the sensor provided by the manufacturer.
Finally, remotely sensed temperatures are influenced by the environmental
conditions present at the time of UAV flight. Atmospheric correction of
surface temperature is essential to extract absolute temperature
measurements from thermal images, requiring the application of a
radiative transfer model [41], a vicarious calibration [42], or an empirical
method [43]. In this research, using an empirical method, two extreme
temperature panels of 0.5 × 0.5m were placed in the plot to record the
hottest (black polymer panel) and the coldest (white polymer panel)
temperatures on scene. Reference panels were close to five times larger than
the GSD and, therefore, several homogeneous pixels appear in the thermal
images. A temperature measurement of each reference panel was collected
Capítulo 1
pág. 55
for each UAV flight with a Flir E60 heat gun (Flir Systems, Oregon, USA).
These measurements were later used to correct the atmospheric effect on
the thermal mosaics, applying an empirical line method, which defines a
linear relationship between absolute temperature and sensor temperature
[44].
2.3. Validation
The analysis of the proposed methodology was applied to all UAV
flights both with and without thermal drift correction. A total of 37
georeferenced checkpoints (CPs) were placed along a transect
perpendicular to the laps and crossing the middle zone of the plot and were
read for temperature using a Flir E60 heat gun. These values were
compared to the extracted values from the thermal orthomosaics obtained
from the flights. The mean error and root mean square error (RMSE) were
calculated for each model and flight. Moreover, Akaike’s information
criterion (AIC) [45] was used to identify the relative importance among all
possible sets of DCMs per UAV flight where the best performing in each
flight was identified by the lowest AIC score.
In addition, the correlation between flight time and thermal drift was
defined. To do this, the thermal images in which a CP appeared were
analyzed. Of all the possible thermal images where a CP appeared, the
images with the most centrally situated CP were used in the mosaic process
to obtain a thermal orthomosaic, as this is the standard method. As each
image has an associated timestamp, each CP was temporally registered as
it appeared in flight, which was then evaluated for the influence of flight
time on thermal drift for each thermal orthomosaic obtained by a DCM.
3. Results
Figure 4 shows the variation of digital number per second (∆𝐷𝑁/𝑡)
along the duration of the UAV flights on features extracted from images
and the different DCMs per flight developed for the proposal methodology.
Comparing the four UAV flights, digital number per second in raw images
varied both in absolute and relative terms. Therefore, the thermal sensor
did not show a defined pattern in registering temperatures. Each DCM
calculated per flight and its correlation coefficient (𝑅2) are summarized in
Table 2. At 8:30 a.m. (Figure 4a), ∆𝐷𝑁/𝑡 had no defined behavior, showing
a disperse distribution. Regardless of the mathematical model used, the
correlation coefficient showed a value between 0.370 with the linear model
and 0.382 for the exponential model of order 2. Therefore, there were no
clear differences a priori between DCMs applied to thermal orthomosaics.
Capítulo 1
pág. 56
At 12.30 p.m. (Figure 4b), ∆𝐷𝑁/𝑡 started to show a trend in its relation to
flight duration as 𝑅2 of some DCMs revealed. In this case, at the beginning
of the UAV flight the ∆𝐷𝑁/𝑡 showed lower values over time (between 0
and 100 s) but then the variation increased, reaching the highest variation
between 200 and 400 s. Thereafter, the variation decreased to values equal
to the beginning of the UAV flight until 700 s where it again increased,
although it did not reach the initial maximum values. In this flight, the
bicubic model showed the highest 𝑅2, equal to 0.482, while the exponential
model had the lowest, 0.190. At 16.00 p.m. (Figure 4c), ∆𝐷𝑁/𝑡 showed a
defined evolution as the UAV flight progressed, which was reflected in
higher 𝑅2 values for all the DCMs. In this mission, ∆𝐷𝑁/𝑠𝑒𝑐 showed
maximum values at the beginning of the UAV flight and then decreased
over time. The exponential order 2 and lineal models had the lowest 𝑅2
value (0.688) while the bicubic model had, again, the highest 𝑅2 value
(0.836). The other DCMs had an 𝑅2 value higher than 0.7. Finally, the
mission at 18:30 p.m. (Figure 4d) showed more oscillations during flight
time; however, the range of ∆𝐷𝑁/𝑠𝑒𝑐 values was narrow. With this mission,
maximum variations occurred at the beginning of the UAV flight,
decreasing as the flight progressed. 𝑅2 values were between 0.67 and 0.69
with no clear differences between the DCMs. Therefore, in analyzing the
behavior of ∆𝐷𝑁/𝑡 along the UAV flight duration for these missions, it is
possible to assert that the thermal sensor was not stable and that its
operation varied in each UAV flight.
Regarding 𝑅2, the DCMs obtained from the UAV flights at 16:00 p.m.
and 18:30 p.m. had higher values than those from 8:30 a.m. and 12:30 p.m.
because ∆𝐷𝑁/𝑠𝑒𝑐 showed a shorter range of values by time. This is due to
the fact that ∆𝐷𝑁/𝑡 per second was different for each UAV flight, occurring
with a lower frequency at later UAV flights.
Capítulo 1
pág. 57
Figure 4. Relationship obtained between the variation of digital number
per second (∆DN/sec) and flight duration for exponential, exponential
order 2, lineal, quadratic, bicubic, and quartic drift correction models at
(a) 8:30 a.m., (b) 12:30 p.m., (c) 16:00 p.m., and (d) 18:00 p.m.
Capítulo 1
pág. 58
Table 2. Drift correction models for each mission and coefficient
of correlation (𝐑𝟐).
DCM
Type
Tim
e of
Flig
ht
Equation 𝐑𝟐
Exponen
tial order
1
8:30 c = 7.160 ∙ e−5.677∙10−4∙t 0.375 *
12:0
0 c = 9.317 ∙ e−4.1911∙10−4∙t
0.190
n.s.
16:0
0 c = 19.923 ∙ e−0.001∙t
0.733
**
18:3
0 c = 18.368 ∙ e−0.001∙t
0.688
**
Exponen
tial order
2
8:30 c = 3.833 ∙ 10−7 ∙ e0.017∙t + 7.227 ∙ e−6.131∙10−4∙t 0.382 *
12:0
0 c = 13.121 ∙ e−9.923∗10−4∙t − 6.958 ∙ e−0.007∙t 0.398 *
16:0
0 c = 20.316 ∙ e−0.001∙t + 8.155 ∙ 10−14 ∙ e−0.36∙t
0.688
**
18:3
0 c = 6.985 ∙ 104 ∙ e−2.791∙t − 6.9833 ∙ 10−4 ∙ e−2.7889∙10−4∙t
0.675
**
Lineal
8:30 c = −0.003 ∙ t + 7.048 0.370 *
12:0
0 c = −0.003 ∙ t + 9.321
0.210
n.s.
16:0
0 c = −0.015 ∙ t + 18.062
0.688
**
18:3
0 c = −0.015 ∙ t + 16.991
0.673
**
Quadrati
c
8:30 c = 2.339 ∙ 10−6 ∙ t2 − 0.005 ∙ t + 7.28 0.379 *
12:0
0 c = −9.601 ∙ 10−6 ∙ t2 + 0.004 ∙ t + 8.120
0.286
n.s.
16:0
0 c = −2.117 ∙ 10−5 ∙ t2 + 0.034 ∙ t + 20.780
0.755
**
18:3
0 c = −4.587 ∙ 10−6 ∙ t2 + 0.019 ∙ t + 17.576
0.676
**
Bicubic
8:30 c = 4.786 ∙ 10−9 ∙ t3 − 3.401 ∙ t2 − 0.003 ∙ t + 7.117 0.380 *
12:0
0 c = 6.913 ∙ 10−8 ∙ t3 − 9.654 ∙ 10−5 ∙ t2 + 0.033 ∙ t + 6.098 0.482 *
16:0
0 c = 7.632 ∙ 10−8 ∙ t3 − 7.838 ∙ 10−5 ∙ t2 + 4.128 ∙ 10−4 ∙ t + 18.299
0.836
**
18:3
0 c = 4.006 ∙ 10−8 ∙ t3 − 4.858 ∙ 10−5 ∙ t2 − 0.001 ∙ t + 16.233
0.694
**
Quartic
8:30 c = 1.630 ∙ 10−11 ∙ t4 − 2.185 ∙ t3 + 1.056 ∙ 10−6 ∙ t2 − 0.005 ∙ t
+ 7.270 0.381 *
12:0
0 c = 7.068 ∙ 10−11 ∙ t4 − 4.838 ∙ 10−8 ∙ t3 − 3.426 ∙ 10−5 ∙ t2 + 0.002
∙ t + 6.51 0.474 *
16:0
0 c = −3.173 ∙ 10−6 ∙ t4 + 6.261 ∙ 10−7 ∙ t3 − 3.828 ∙ 10−4 ∙ t2
+ 0.058 ∙ t + 15.927 0.799
**
18:3
0 c = −1.351 ∙ 10−10 ∙ t4 + 2.791 ∙ 10−7 ∙ t3 − 1.837 ∙ 10−4 ∙ t2
+ 0.025 ∙ t + 15.138 0.688
**
Capítulo 1
pág. 59
The asterisks indicate the level of significance (* p < 0.05, ** p < 0.001,
n.s. not significant).
Figure 5 shows the thermal orthomosaic histograms for each flight
mission with the applied DCMs, as well as without a DCM; Table 3
summarizes the statistics for each. First, the histograms manifest a clear
difference between the thermal orthomosaics where a DCM was applied
compared to those where no DCM was applied. Moreover, the temperature
distribution was quite similar in all the thermal orthomosaics where a DCM
was applied. Because the study area had two differentiated classes,
vegetation and bare soil, a bimodal distribution was expected to describe
temperature distribution of the scene. However, at the 8:30 a.m. UAV
mission (Figure 5a), all thermal orthomosaics had a normal distribution as
Sarle’s bimodality coefficient (SBC) showed. The temperature range on
those thermal orthomosaics where a DCM was applied ranged from 15 to
35 °C, 20 °C occurring with the most frequency. Conversely, the thermal
orthomosaic without any drift correction showed a broader temperature
range from 15 to 43 °C and a right-skewed distribution. Instead of a
bimodal distribution, a normal distribution of temperatures occurred in the
early morning on the thermal orthomosaics with drift correction because
the bare soil and vegetation had not yet absorbed heat from the sunlight
and consequently the temperatures of both were similar. Therefore, at this
hour, both classes showed no clear difference in thermal behavior.
Conversely, at 12:30 p.m. (Figure 5b), the UAV flight histograms showed
two different shapes irrespective of whether a DCM was applied or not.
However, when no drift correction was applied, the temperature
distribution was, again, similar to a normal distribution with SBC being less
than 5/9. As a result, the non-corrected histogram did not properly mark
bare soil and vegetation with different temperatures. On the other hand, all
drift-corrected thermal orthomosaics showed an SBC higher than 5/9,
having a bimodal distribution with two differentiated peaks. In this
mission, bare soil and vegetation had different behaviors as they correlated
to sunlight. Although both classes increased their temperature, vegetation
(left peak) showed a mean temperature equal to 30 °C, which was lower
than bare soil (right peak), which reached a mean value equal to 50 °C. At
16:30 p.m. (Figure 5c), both vegetation and bare soil increased in
temperature, which was properly shown when a DCM was applied. If
corrections were not applied, the temperatures were also higher but the
classes were not accurately represented in the histogram. A comparison of
the histograms shows that vegetation temperature increased as the day
progressed, reaching the highest temperature at 18:30 p.m. while soil
temperature increased until 16:30 p.m. and then began to decrease. At 18:30
Capítulo 1
pág. 60
p.m. (Figure 5d), the left peak of the histogram has a higher frequency than
the right peak as the vegetation maintained a stable temperature while the
bare soil temperature decreased, which also caused the distance between
the peaks to be reduced. These results are because the soil was cooling due
to the declining sun and greater shadow cover.
All of the DCMs used successfully described this occurrence from
12:30 p.m. to 18:30 p.m. Moreover, the histograms of each DCM for every
flight had a similar distribution with the quartic model presenting the
greatest differences in portions of the flights. The histograms show that
from 12:30 p.m., the temperature difference between the vegetation and
bare soil was quite distinctive and remained so until 16:30 p.m. when the
difference began to decrease. Therefore, as in Bellvert et al. (2014) [43], it is
recommended that thermal UAV flights with agronomic objectives are
performed between 12:00 and 16:30 p.m. However, without the method
proposed in this paper, non-corrected thermal orthomosaics will not
sufficiently differentiate soil and vegetation.
Capítulo 1
pág. 61
Figure 5. Thermal orthomosaic histograms at (a) 8:30 a.m., (b) 12:30 p.m.,
(c) 16:30 p.m., and (d) 18:30 p.m. for each drift correction model and
without correction.
Capítulo 1
pág. 62
Table 3. Statistics of thermal orthomosaics.
Time of
Flight Expone
ntial
Exponential
Order 2
Lin
eal
Quad
ratic
Cu
bic
Qua
rtic
No
DCM
8:30 a.m. Ran
ge 35.33 35.66
35.6
6 35.68
34.
64
34.3
3 43.16
Me
an 20.92 20.96
20.8
8 20.95
20.
96
21.1
0 28.35
SD 3.29 3.28 3.30 3.29 3.2
8 3.26 7.14
SB
C 0.38 0.38 0.38 0.38
0.3
8 0.38 0.47
12:30
p.m.
Ran
ge 50.08 50.03
50.1
5 50.31
49.
80
50.7
2 62.38
Me
an 40.74 40.50
40.5
3 40.48
40.
61
40.9
8 47.68
SD 8.87 8.89 8.92 8.91 8.8
7 8.80 10.43
SB
C 0.66 0.66 0.66 0.66
0.6
6 0.67 0.45
16:00
p.m.
Ran
ge 50.89 49.82
50.5
2 50.08
50.
18
48.7
1 74.52
Me
an 49.19 49.20
48.6
6 49.22
49.
37
47.8
4 58.66
SD 9.41 9.24 9.35 9.25 9.1
9 9.73 13.06
SB
C 0.68 0.68 0.68 0.68
0.6
8 0.65 0.48
18:30
p.m.
Ran
ge 43.78 39.65
42.8
3 41.17
42.
46
43.7
4 49.57
Me
an 39.57 39.47
39.3
7 39.46
39.
55
38.6
1 46.69
SD 5.24 5.26 5.28 5.24 5.2
3 5.55 8.32
SB
C 0.67 0.67 0.66 0.67
0.6
7 0.66 0.48
SBC: Sarle’s bimodality coefficient. SD: standard deviation.
Validation
Once the histograms of the drift corrected thermal orthomosaics
accurately described the presence of vegetation and bare soil in the study
Capítulo 1
pág. 63
area, the next step was to analyze which DCM had the greatest thermal
accuracy. Figure 6 illustrates the 16:30 p.m. thermal orthomosaics with the
applied DCMs, which are detailed in Table 2, as well as without drift
correction. When drift correction was not applied (Figure 6g), the resulting
thermal orthomosaics ordered the temperature values. From a visual
analysis, neither bare soil nor vegetation showed stable temperature.
Moreover, temperature changed along the north and south laps, registering
higher temperature values as the UAV flight progressed. This effect is
pronounced in this paper because NUC was switched off to obtain an
extreme example; in other cases, the drift effect would be less pronounced.
This also explains the pronounced skewness in the non-corrected
histograms presented in Figure 5. Based on the authors’ results, the
temperature variation for this sensor under normal conditions was less
than 0.5 °C per minute, similar to the variations reported by Olbrycht and
Więcek (2015) [29].
Regarding the thermal orthomosaics where a DCM was applied
(Figure 6a–f), no visual temperature differences from north to south were
detected. Instead, the temperature variations were linked to the state of
vegetation and bare soil as explained by the histogram analysis above. In
addition, comparing all of the thermal orthomosaics generated with a DCM
resulted in similar orthomosaics with the exception of the quartic model
(Figure 6f). This model generated colder temperatures in the south of the
plot compared to the north of the plot, which was not present in the other
DCMs. These differences were not detected in the field campaign,
suggesting that the quartic model did not adequately describe thermal drift
in the UAV flights. This result occurred for all of the UAV flights assessed.
Table 4 summarizes the results of thermal quality control on the
thermal orthomosaics from each UAV flight with and without the applied
DCMs by mean error and standard deviation (SD) and Akaike’s
information criterion (AIC). In addition, a correlation coefficient (r2)
between error and flight progress was calculated. In all the orthomosaics,
as expected, where a DCM was not applied, the temperature errors were
higher than where a DCM was applied. Therefore, it is necessary to pre-
process thermal images taking into account the behavior of the
microbolometer registering temperature values. The 8:30 a.m. mission
showed a higher rate of error than the other UAV missions independent of
which DCM was applied. At this time, the error ranged from 0.88 ± 0.8 °C
using the bicubic model correction to 1.01 ± 0.81 °C using the exponential
correction model. These higher errors are explained by the different
environmental conditions while performing the UAV flight and measuring
ground truth. In the early morning, the sun is ascending and an object’s or
Capítulo 1
pág. 64
coverage’s superficial temperature changes in a short time interval.
Although ground truth measurements were performed immediately after
the UAV flight finished (twenty minutes), the duration of the UAV flight
was long enough for temperature values of the olive trees and bare soil to
change both for images and for on-ground measurements, influencing this
mission’s results. At 12:30 p.m., the mean error decreased to values around
0.2 °C ± 0.5 °C for all DCMs applied with the cubic model generating the
smallest error (0.10 °C ± 0.45 °C) while the exponential model order 2 had
the highest error (0.29 °C ± 0.56 °C). Conversely, at 16:30 p.m., the errors
differed depending on the DCM used. For this mission, the quartic model
showed the highest error (1.57° ± 1.22 °C) while the cubic model had the
lowest (0.06 ± 0.45 °C). Finally, at 18:30 p.m., the cubic model again showed
the greatest accuracy with an error equal to 0.26 ± 0.58 °C and the quartic
model being the worst with an error equal to 1.55 ± 0.85 °C. Even so, the
highest mean error and SD were found in those thermal orthomosaics
where no DCM was applied, independent of the mission. Moreover, the
calculated AIC scores identified the cubic model as the most consistent
DCM, showing the minimum score in each UAV flight.
Capítulo 1
pág. 65
Figure 6. Comparison of thermal orthomosaics at 16:30 p.m. with drift
model corrections: (a) exponential, (b) exponential order 2, (c) lineal, (d)
quadratic, (e) bicubic, (f) quartic, and (g) no correction.
Table 4. Mean error, standard deviation (SD), Akaike’s
information criterion (AIC), and coefficients of correlation (r2) for
each model correction and mission.
Exponen
tial
Exponenti
al 2
Line
al
Quadra
tic
Cub
ic
Quar
tic
No
DCM
8:3
0
Mea
n 1.013 0.898 0.913 0.901 0.88 0.895 −2.929
SD 0.818 0.804 0.815 0.800 0.80 0.758 2.792
AIC 3.777 0.832 3.593 2.574 0.77 18.47
2 ‐‐
r2 0.027 0.004 0.001 0.005 0.04 0.645
**
0.918
**
12:
30
Mea
n 0.288 0.291 0.225 0.222 0.10 0.212 −3.555
SD 0.590 0.560 0.556 0.569 0.45 0.584 3.012
AIC 39.833 27.437 36.36
4 31.673 26.6
56.95
5 ‐‐
r2 0.007 0.025 0.052 0.08 0.06 0.322
**
0.959
**
16:
30
Mea
n 0.112 ‐0.152 0.518 −0.123 −0.06 1.573
−12.15
8 SD 0.588 0.493 0.795 0.508 0.45 1.224 7.363
AIC 18.219 7.452 11.32
9 7.765 2.31
31.89
3 ‐‐
r2 0.391 ** 0.156 * 0.667
** 0.168 * 0.00
0.359
**
0.915
**
18:
30
Mea
n 0.379 0.409 0.524 0.404 0.26 1.557 −9.544
SD 0.648 0.622 0.627 0.626 0.58 0.857 5.433
AIC 12.992 6.888 13.87
1 9.006 2.56
29.43
2 ‐‐
r2 0.001 0.009 0.032 0.005 0.02 0.568
**
0.880
**
Pearson’s analysis. ** Correlation is significant at the 0.01 level (2-tailed);
* Correlation is significant at the 0.05 level (2-tailed).
Capítulo 1
pág. 66
To demonstrate, Figure 7 shows the relationship between temperature
obtained from the UAV flight and the on-ground measurements at 16:30
p.m. The continuous line represents a perfect correlation of temperatures
between both sets of temperatures and a discontinuous line represents the
adjusted lineal model from both sets. When no DCM was applied (Figure
7g), the temperature values in the thermal orthomosaic did not show any
relationship with those on the ground having a broad range of variation.
On the other hand, in general, when a DCM was applied, temperature
values on the thermal orthomosaics and their corresponding on-ground
measurement were similar. However, there were differences between the
DCMs. The quartic model (Figure 7f) yielded the highest deviation from a
perfect correlation although not as broad as in the case of not using a DCM.
The same occurred when the exponential model (Figure 7a) or lineal model
(Figure 7c) was applied, although not as evident as in the previous case.
The remaining DCMs considered did not show significant differences in
their temperature values.
Capítulo 1
pág. 67
Figure 7. Correlation for each drift correction model (DCM) between
temperature values from the thermal orthomosaic and on-ground
measurement at 16:30 p.m. unmanned aerial vehicle (UAV) flight. DCM:
(a) exponential, (b) exponential order 2, (c) lineal, (d) quadratic, (e) cubic,
(f) quartic and (g) no correction.
Moreover, to have an acceptable range of temperature error, it is
necessary that this error occurs independently of the flight duration,
meaning that the drift effect of the microbolometer has been adjusted
accordingly. Table 4 shows this relationship through a correlation
coefficient and its Pearson analysis and Figure 8 shows the relation between
error and flight duration for each thermal orthomosaic obtained from the
16:30 p.m. mission where a DCM was applied. As expected, the thermal
orthomosaics where a DCM was not applied showed high R2 between error
and flight duration in all UAV flights, meaning that the error is dependent
Capítulo 1
pág. 68
on flight duration. As to the thermal orthomosaics where a DCM was
applied, the quartic model was the one whose errors showed a significant
correlation with flight duration in all UAV flights. The other DCMs had no
significant correlation at 8:30 a.m., 12:30 p.m., and 18:30 p.m. At 16:30 p.m.,
the cubic model was the only DCM whose results were independent of
flight duration. Therefore, when considering the relationship between
temperature error and flight duration, the cubic model adequately
described drift effect on temperature measurements on all performed UAV
flights. In this study and our experience, using the Gobi 640 thermal sensor
on subsequent UAV flights has shown that the cubic drift model is the most
adequate. However, for other thermal sensors, it is recommendable to
analyze which model would better describe drift effect using the proposed
methodology.
Figure 8. Relationship for each DCM between temperature error and
flight duration for the 16:30 p.m. mission. DCM: (a) exponential, (b)
exponential order 2, (c) lineal, (d) quadratic, (e) cubic, (f) quartic.
The temperature error obtained in this research when a DCM was
applied is equal to those described in the literature [8,38] and therefore
presents itself as another option for agronomical projects. Moreover, the
proposed methodology presents an advantage as it optimizes flight time.
In other studies, it is necessary to stabilize the sensor in flight [36] or it is
necessary to recurrently fly over temperature targets for reference to
calibrate thermal images [38]. These two strategies spend the limited
Capítulo 1
pág. 69
battery charge and, therefore, reduce the area covered by the UAV flight
due to the decrease in useful flight time. With the proposed methodology,
flight time is maximized without loss of accuracy in temperature values.
Moreover, it allows the use of a thermal sensor on board the UAV
regardless of the drift effect.
The UAV flights were performed under stable weather conditions and
were of short duration (20 min) and as such, there were no changes in the
surface temperature. However, UAV flights with longer duration (longer
than 30 min) and/or under unstable weather conditions must be evaluated
in future work due to changes in atmospheric or sunlight conditions. In
these cases, in addition to the drift correction, it would be necessary to have
an instantaneous atmospheric correction adapted to the UAV flight
conditions. One possible methodology is to equip the UAV platform with
additional sensors that record values of temperature, relative humidity,
atmospheric pressure, incident radiation, and wind speed. These
parameters would allow the possibility of determining the atmospheric
conditions linked to each individual thermal image and, therefore, at each
moment of flight time. These parameters along with the drift effect model
would allow more precise and accurate temperature values.
4. Conclusions
Remote sensing using lightweight uncooled thermal sensors on board
UAVs is a useful tool for measuring crop temperatures. However, drift
effect on registered temperatures can invalidate its agronomical
applications where an accuracy greater than 1 °C is necessary. The present
research has developed methodology to remove drift effect on temperature
using a lightweight microbolometer thermal sensor on board a UAV. In this
study, removing drift effect on thermal images is based on redundant
information around objects that appear in different overlapping images
from a UAV flight that covers the area of interest. Different mathematical
models were explored to describe drift effect with the cubic drift model
yielding the best results on separate missions performed for this research.
These models were tested in four UAV missions at different hours at
the same location. If no drift correction was applied, the thermal
orthomosaics did not adequately describe crop temperatures, invalidating
their use within an agronomical context. Contrarily, if a drift correction
model was applied using the proposed methodology, the results improved
with a range of error that would be adequate for agronomical projects.
Moreover, the accuracy is in the same range as other authors’ results but
with the added benefit of not requiring any special in air UAV flight
Capítulo 1
pág. 70
operation and thusly increasing useful flight time. This is an important
point to those UASs with short flight durations due to limited battery
power. In addition, this methodology was applied to a single UAV flight
and, therefore, the proposed drift correction model is adaptable to specific
flight conditions.
In this study, the cubic drift model offered the best results. However,
the authors recommend exploring the behavior of a particular uncooled
thermal sensor to determine which model would best describe the drift
effect using the proposed methodology.
Acknowledgments: This research was funded by Khepri project supported by
CDTI: Centro para el Desarrollo Tecnológico Industrial, FEDER Funds: Fondo
Europeo de Desarrollo Regional and CTA: Corporación Tecnológica de Andalucía.
Author Contributions: Francisco Javier Mesas-Carrascosa and Fernando Pérez
Porras. conceived and designed the experiments; Francisco Javier Mesas-
Carrascosa, Fernando Pérez Porras, Jose Emilio Meroño de Larriva. and Alfonso
García-Ferrer performed the experiments; Francisco Javier Mesas-Carrascosa,
Fernando Pérez Porras, Carlos Mena Frau, Francisco Agüera-Vega, Fernando
Carvajal-Ramírez. and Patricio Martinez-Carricondo analyzed the data; Francisco
Javier Mesas-Carrascosa wrote the paper.
Conflicts of Interest: The authors declare no conflict interest.
References
1. Alexandratos, N.; Bruinsma, J. World Agriculture Towards 2030/2050: The
2012 Revision; Agriculture Development Economics Division Food and
Agriculture Organization of the United Nations: 2012. Publisher: Publication
(Rome, Italy). Avaliable online:
http://www.fao.org/docrep/016/ap106e/ap106e.pdf (accessed on 16th April
2018.)
2. McCalla, A.F. Challenges to world agriculture in the 21st century. UPDATE
Agric. Resour. Econ. 2001, 4. Available online:
https://giannini.ucop.edu/publications/are-
update/issues/2001/4/3/challenges-to-world-agric/ (accessed on 16th April
2018)
3. Stafford, J.V. Implementing precision agriculture in the 21st century. J. Agric.
Eng. Res. 2000, 76, 267–275.
4. Miao, Y.; Mulla, D.J.; Randall, G.W.; Vetsch, J.A.; Vintila, R. Combining
chlorophyll meter readings and high spatial resolution remote sensing images
for in-season site-specific nitrogen management of corn. Precis. Agric. 2009,
10, 45–62.
5. Mulla, D.J. Twenty five years of remote sensing in precision agriculture: Key
advances and remaining knowledge gaps. Biosyst. Eng. 2013, 114, 358–371.
Capítulo 1
pág. 71
6. Goetz, A.F.H. Three decades of hyperspectral remote sensing of the earth: A
personal view. Remote Sens. Environ. 2009, 113, S5–S16.
7. Anderson, K.; Gaston, K.J. Lightweight unmanned aerial vehicles will
revolutionize spatial ecology. Front. Ecol. Environ. 2013, 11, 138–146.
8. Berni, J.A.J.; Zarco-Tejada, P.J.; Sepulcre-Cantó, G.; Fereres, E.; Villalobos, F.
Mapping canopy conductance and cwsi in olive orchards using high
resolution thermal remote sensing imagery. Remote Sens. Environ. 2009, 113,
2380–2388.
9. Colomina, I.; Molina, P. Unmanned aerial systems for photogrammetry and
remote sensing: A review. ISPRS J. Photogramm. Remote Sens. 2014, 92, 79–
97.
10. McBratney, A.; Whelan, B.; Ancev, T.; Bouma, J. Future directions of precision
agriculture. Precis. Agric. 2005, 6, 7–23.
11. Lin, Y.; Hyyppa, J.; Jaakkola, A. Mini-uav-borne lidar for fine-scale mapping.
IEEE Geosci. Remote Sens. Lett. 2011, 8, 426–430.
12. Wallace, L.; Lucieer, A.; Watson, C.; Turner, D. Development of a uav-lidar
system with application to forest inventory. Remote Sens. 2012, 4, 1519–1543.
13. Torres-Sánchez, J.; Peña, J.M.; de Castro, A.I.; López-Granados, F. Multi-
temporal mapping of the vegetation fraction in early-season wheat fields
using images from uav. Comput. Electron. Agric. 2014, 103, 104–113.
14. López-Granados, F.; Torres-Sánchez, J.; Serrano-Pérez, A.; de Castro, A.I.;
Mesas-Carrascosa, F.-J.; Peña, J.-M. Early season weed mapping in sunflower
using uav technology: Variability of herbicide treatment maps against weed
thresholds. Precis. Agric. 2016, 17, 183–199.
15. Mesas-Carrascosa, F.-J.; Torres-Sánchez, J.; Clavero-Rumbao, I.; García-Ferrer,
A.; Peña, J.-M.; Borra-Serrano, I.; López-Granados, F. Assessing optimal flight
parameters for generating accurate multispectral orthomosaicks by uav to
support site-specific crop management. Remote Sens. 2015, 7, 12793–12814.
16. Candiago, S.; Remondino, F.; De Giglio, M.; Dubbini, M.; Gattelli, M.
Evaluating multispectral images and vegetation indices for precision farming
applications from uav images. Remote Sens. 2015, 7, 4026–4047.
17. Zarco-Tejada, P.J.; Guillén-Climent, M.L.; Hernández-Clemente, R.; Catalina,
A.; González, M.R.; Martín, P. Estimating leaf carotenoid content in vineyards
using high resolution hyperspectral imagery acquired from an unmanned
aerial vehicle (uav). Agric. Forest Meteorol. 2013, 171–172, 281–294.
18. Hruska, R.; Mitchell, J.; Anderson, M.; Glenn, N.F. Radiometric and geometric
analysis of hyperspectral imagery acquired from an unmanned aerial vehicle.
Remote Sens. 2012, 4, 2736–2752.
19. Baluja, J.; Diago, M.P.; Balda, P.; Zorer, R.; Meggio, F.; Morales, F.; Tardaguila,
J. Assessment of vineyard water status variability by thermal and
multispectral imagery using an unmanned aerial vehicle (uav). Irrig. Sci. 2012,
30, 511–522.
Capítulo 1
pág. 72
20. Gonzalez-Dugo, V.; Zarco-Tejada, P.; Nicolás, E.; Nortes, P.A.; Alarcón, J.J.;
Intrigliolo, D.S.; Fereres, E. Using high resolution uav thermal imagery to
assess the variability in the water status of five fruit tree species within a
commercial orchard. Precis. Agric. 2013, 14, 660–678.
21. Fereres, E.; Soriano, M.A. Deficit irrigation for reducing agricultural water use.
J. Exp. Bot. 2007, 58, 147–159.
22. Calderón, R.; Navas-Cortés, J.A.; Lucena, C.; Zarco-Tejada, P.J. High-
resolution airborne hyperspectral and thermal imagery for early detection of
verticillium wilt of olive using fluorescence, temperature and narrow-band
spectral indices. Remote Sens. Environ. 2013, 139, 231–245.
23. Zarco-Tejada, P.J.; González-Dugo, V.; Berni, J.A.J. Fluorescence, temperature
and narrow-band indices acquired from a uav platform for water stress
detection using a micro-hyperspectral imager and a thermal camera. Remote
Sens. Environ. 2012, 117, 322–337.
24. Gonzalez-Dugo, V.; Hernandez, P.; Solis, I.; Zarco-Tejada, P.J. Using high-
resolution hyperspectral and thermal airborne imagery to assess physiological
condition in the context of wheat phenotyping. Remote Sens. 2015, 7, 13586–
13605.
25. Chapman, S.C.; Merz, T.; Chan, A.; Jackway, P.; Hrabar, S.; Dreccer, M.F.;
Holland, E.; Zheng, B.; Ling, T.J.; Jimenez-Berni, J. Pheno-copter: A low-
altitude, autonomous remote-sensing robotic helicopter for high-throughput
field-based phenotyping. Agronomy 2014, 4, 279–301.
26. Gallo, M.A.; Willits, D.S.; Lubke, R.A.; Thiede, E.C. Low-cost uncooled ir
sensor for battlefield surveillance. Proceed. SPIE 1993, 2020, pp 351–362.
27. Krupiński, M.; Bareła, J.; Firmanty, K.; Kastek, M. Test stand for non-
uniformity correction of microbolometer focal plane arrays used in thermal
cameras. Proceed. SPIE Int. Soc. Opt. Eng. 2013, 8896, doi:10.1117/12.2028633.
28. Huawei, W.; Caiwen, M.; Jianzhong, C.; Haifeng, Z. An adaptive two-point
non-uniformity correction algorithm based on shutter and its implementation.
In Proceedings of the 2013 Fifth International Conference on Measuring
Technology and Mechatronics Automation, Hong Kong, China, 16–17 January
2013; pp 174–177.
29. Olbrycht, R.; Więcek, B. New approach to thermal drift correction in
microbolometer thermal cameras. Quant. InfraRed Thermogr. J. 2015, 12, 184–
195.
30. Mudau, A.E.; Willers, C.J.; Griffith, D.; Roux, F.P.J.L. Non-uniformity
correction and bad pixel replacement on lwir and mwir images. In
Proceedings of the 2011 Saudi International Electronics, Communications and
Photonics Conference (SIECPC), Riyadh, Saudi Arabia, 24–26 April 2011; pp
1–5.
31. King, S.R.; Rekow, M.N.; Carlson, P.S.; Heinke, T.; Warnke, S.H.; Brest, B.
Shutterless Infrared Imager Algorithm with Drift Correction. Google Patents
US8067735B2, 2011.
Capítulo 1
pág. 73
32. Tempelhahn, A.; Budzier, H.; Krause, V.; Gerlach, G. Shutter-less calibration
of uncooled infrared cameras. J. Sens. Sensor Syst. 2016, 5, doi:10.5194/jsss-5-
9-2016.
33. Mizrahi, U.; Fraenkel, A.; Kopolovich, Z.; Adin, A.; Bikov, L. Method and
System for Measuring and Compensating for the Case Temperature Variations
in a Bolometer Based System. US Patent No. US 7807968 B2, 2010.
34. Harris, J.G.; Yu-Ming, C. Nonuniformity correction of infrared image
sequences using the constant-statistics constraint. IEEE Trans. Image Process.
1999, 8, 1148–1151.
35. Zuo, C.; Chen, Q.; Gu, G.; Sui, X.; Ren, J. Improved interframe registration
based nonuniformity correction for focal plane arrays. Infrared Phys. Technol.
2012, 55, 263–269.
36. Berni, J.A.J.; Zarco-Tejada, P.J.; Suarez, L.; Fereres, E. Thermal and
narrowband multispectral remote sensing for vegetation monitoring from an
unmanned aerial vehicle. IEEE Trans. Geosci. Remote Sens. 2009, 47, 722–738.
37. Zalameda, J.N.; Winfree, W.P. Investigation of uncooled microbolometer focal
plane array infrared camera for quantitative thermography. J. Nondestruct.
Eval. 2005, 24, 1–9.
38. Gómez-Candón, D.; Virlet, N.; Labbé, S.; Jolivot, A.; Regnard, J.-L. Field
phenotyping of water stress at tree scale by uav-sensed imagery: New insights
for thermal acquisition and calibration. Precis. Agric. 2016, 17, 786–800.
39. Jensen, A.M.; McKee, M.; Chen, Y. Procedures for processing thermal images
using low-cost microbolometer cameras for small unmanned aerial systems.
In Proceedings of the 2014 IEEE Geoscience and Remote Sensing Symposium,
Quebec City, QC, Canada, 13–18 July 2014; pp 2629–2632.
40. Lowe, D.G. Distinctive image features from scale-invariant keypoints. Int. J.
Comput. Vis. 2004, 60, 91–110.
41. Zarco-Tejada, P.J.; Berni, J.A.J.; Suárez, L.; Sepulcre-Cantó, G.; Morales, F.;
Miller, J.R. Imaging chlorophyll fluorescence with an airborne narrow-band
multispectral camera for vegetation stress detection. Remote Sens. Environ.
2009, 113, 1262–1275.
42. Torres-Rua, A. Vicarious calibration of suas microbolometer temperature
imagery for estimation of radiometric land surface temperature. Sensors 2017,
17, 1499.
43. Honkavaara, E.; Saari, H.; Kaivosoja, J.; Pölönen, I.; Hakala, T.; Litkey, P.;
Mäkynen, J.; Pesonen, L. Processing and assessment of spectrometric,
stereoscopic imagery collected using a lightweight uav spectral camera for
precision agriculture. Remote Sens. 2013, 5, 5006–5039.
44. Bellvert, J.; Zarco-Tejada, P.J.; Girona, J.; Fereres, E. Mapping crop water stress
index in a ‘pinot-noir’ vineyard: Comparing ground measurements with
thermal remote sensing imagery from an unmanned aerial vehicle. Precis.
Agric. 2014, 15, 361–376.
Capítulo 1
pág. 74
45. Akaike, H. A new look at the statistical model identification. IEEE Trans.
Autom. Control 1974, 19, 716–723.
© 2018 by the authors. Submitted for possible open access
publication under the
terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
Capítulo 2
pág. 75
5 CAPÍTULO 2
Publicado en: Remote Sensing, 2019, 11(20), 2413
https://doi.org/10.3390/rs11202413
Project-Based Learning Applied to
Unmanned Aerial Systems and
Remote Sensing
Francisco-Javier Mesas-Carrascosa 1,*, Fernando Pérez Porras 1, Paula
Triviño-Tarradas1, Jose Emilio Meroño de Larriva 1 and Alfonso García-
Ferrer 1
1 Department of Graphic Engineering and Geomatics, University of
Cordoba, Campus de Rabanales, 14071 Córdoba, Spain;
[email protected](F.P.P.); [email protected](P.T-T.); [email protected](J.E.M.);
[email protected](A.G-F.)
* Correspondence: [email protected]
Received: 19 September 2019; Accepted: 16 October 2019; Published: date
Abstract: The development of unmanned aerial vehicle (UAV) technology
and the miniaturization of sensors have changed the way remote sensing
(RS) is used, popularizing this geoscientific discipline in other fields, such
as precision agriculture. This makes it necessary to implement the use of
these technologies in teaching RS alongside the classical platforms
(satellite and manned aircraft). This manuscript describes how The
Higher Technical School of Agricultural Engineering at the University of
Córdoba (Spain) has introduced UAV RS into the academic program by
way of project-based learning (PBL). It also presents the basic
characteristics of PBL, the design of the subject, the description of the
teacher-guided and self-directed activities, as well as the degree of student
satisfaction. The teaching and learning objectives of the subject are to learn
how to determine the vigor, temperature, and water stress of a crop
through the use of RGB, multispectral, and thermographic sensors
onboard a UAV platform. From the onset, students are motivated, actively
participate in the tasks related to the realization of UAV flights, and
subsequent processing and analysis of the registered images. Students
report that PBL is more engaging and allows them to develop a better
understanding of RS.
Keywords: educational assessment; motivation; unmanned aerial vehicle;
agriculture
Capítulo 2
pág. 78
1. Introduction
Universities are tasked with improving student learning and
demonstrating program effectiveness. Even though numerous teaching
tools are available, teaching and learning are not singularly dependent on
technology. However, as technology enhances student access to resources,
we may witness a portentous shift in their learning patterns [1]. One of the
primary components of effective teaching, and likewise critical to learning,
is student engagement [2,3]. In Europe, as defined by the new paradigm of
European Higher Education within the Bologna process, lecturers are
agents that create working environments to stimulate students. In this
scenario, lecturers should create and generate resources to facilitate an
adequate context for learning [4]. The main factor in the learning process is
the willingness to learn [5], which is dependent on the student. However,
the lecturer can help students by guiding and supporting their autonomous
learning. One major objective is to realize the vision of “learning to learn”,
in which students are trained in methods to acquire information, critical
thinking, and life-long learning skills.
Remote sensing (RS) is a rapidly growing technology integrated with
other disciplines, such as photogrammetry, geographic information
systems (GIS), and computer science. New Earth observation programs like
Copernicus [6], acquisition platforms like unmanned aerial systems (UAS)
or cloud-based platforms for geospatial analysis like Google Earth Engine
[7], have placed the geoscientific community, and in particular RS, in an
essential position. Its importance has been recognized by different
organizations [8,9] and has been identified as one of the three most
important emerging disciplines [10]. Geosciences are linked to technologies
related to the Earth’s surface, with GIS, RS, and Global Navigation Systems
(GNSS) being of key import. These technologies are joined under the term
geomatics, defined as the branch of science that deals with the collection,
analysis, and interpretation of data, especially instrumental data, relating
to the Earth’s surface.
The contents and methodologies used in educational programs for
remote sensing have changed radically. At the beginning of the 1980s, most
of the courses offered in the United States were provided as service courses
at an introductory level [11]. However, in 2013, a review of technical
education in Europe on aerospace and RS was carried out and [12]
Capítulo 2
pág. 79
education programs were organized into three main categories: a)
“Aerospace hardware”, b) “Remote sensing and image/data processing”,
and c) “Aerospace environmental application”. The first category is
focused on aircraft or satellite structure as the main objective to study
construction, propulsion, motion, aerodynamics, and flight mechanics of
the platforms. Included in “remote sensing and image/data processing”,
authors highlight programs related to data capture and processing as the
main topics for RS. These are classical courses focused on the optical and
radar sensors and satellite principles and the main image processing
techniques. Additionally, in some cases, the educational scope is
multimedia data processing, with RS being a possible application. Finally,
“Aerospace environmental applications” are focused on RS for
environmental purposes. The use of satellite images to create maps to
monitor forests, oceans, glaciers or urban areas are some of the typical
topics of courses like this. Independent of the geographic area, new
technological developments in RS and photogrammetry areas demand
changes in the educational programs, requiring adaptation and application
of improved teaching methods [13]. Active learning methods have been
proved to better motivate students, increasing their knowledge compared
to traditional ways of teaching [14,15]. Active learning involves students in
doing things and thinking about what they are doing [16]. In this
framework, different approaches were adopted in the remote sensing field
like podcasts created by students [17], inquiry-based educational
experiments [18], internet-based seminars [19], and interactive online tools
[20]. A successful inquiry-based educational experiment is “ESF goes to
space” (College of Environmental Science and Forestry, New York, USA),
where students design, build, and launch a helium balloon for acquiring
images [18].
Commercial opportunities through UAS are on the rise as platform
and sensor technologies are becoming more affordable. This fact is reflected
in the increase of bibliographic references in recent years. Different terms
fall under UAS: Unmanned aerial vehicle (UAV), remotely piloted aircraft
systems (RPAS) or drone [21]. Bibliometric census using the scientific
database Scopus (Figure 1) shows an increase in the number of appearances
year by year (up to 15 April 2019), with UAV being the term most used.
This trend also appears when adding the keyword “agriculture” to this
analysis, showing the interest of its use in agronomy. Equally of note,
precision agriculture (PA) is one of the most economically important
sectors in the UAS market [22,23]. As such, the development of UAS in the
last decade has marked a new era in RS and has become a serious “game-
Capítulo 2
pág. 80
changer” in PA [24], providing data of unprecedented spatial, spectral, and
temporal resolution [21].
Figure 1. Cumulative of UAS technology items added to Scopus from
2014 to 2019: (a) General search and (b) Agriculture search.
Given this scenario, universities should expand their education
programs, including UAV-based geomatics operations and application
developments [25]. New curriculums are mainly focused, to date, on
photogrammetry applications [26,27]. As a result of the interest generated
by UAV applications, it is likely beneficial to introduce UAV-RS learning
in those university studies related to agriculture. The goal of this
manuscript is to show the teaching methodology used to do so in the
Faculty of Agriculture and Forestry Engineering (Escuela Técnica Superior
de Ingeniería Agronómica y de Montes, ETSIAM) at the University of
Córdoba (Spain).
2. Project-Based Learning: Characteristics and Goals
One of the main objectives of engineering programs is “to produce
broad-based, flexible graduates who can think integratively, solve
problems, and be life-long learners” [28]. As such, it is accepted that if a
combination of theory and practice are implemented into the educational
programs, the educational outputs are positive for the student’s learning
process. Learning methods, like project-based learning (PBL), are adequate
supplements to traditional teaching methods. PBL is considered an
approach to teaching where students respond to real-world questions [29],
providing a context of learning through problems or questions linked to
Capítulo 2
pág. 81
real-world practices [30], and dealing with problems that approximate real
situations [31]. Therefore, the main goal of PBL is to provide students the
opportunities to apply knowledge instead of simply acquiring it, with a
focus on problem formulation as well as problem-solving [32]. In this
context, the practice of PBL is supported by three major principles in
learning:
The constructive process: Aiding the processing of new information
and developing connections to previous learning is a basic requirement
for teaching and learning [33]. PBL supports learning as an act of
discovery as students assess the problem, research its background,
analyze a possible solution, develop a proposal, and finally generate a
final result [31].
Metacognition: The monitoring and control of thought [34]. With PBL,
students are able to detect when they understand new information or
not. Therefore, PBL provides the opportunity to monitor and evaluate
the learning progress.
Social and cultural factors: Learning focused on real-world context.
PBL is focused on articulating problems and solving them by
simulating real-world RS research and development.
As per Barrows [35], PBL characteristics are: “a) learning is student-
centered, b) learning occurs in small student groups, c) lecturers are
facilitators or guides, d) problems form the organizing focus and stimulus
for learning, e) problems are a vehicle for the development of clinical
problem-solving skills and f) new information is acquired through self-
directed learning”. PBL is a learning model where students actively work,
plan, implement, and evaluate projects that have real-world applications
[36], which is not to be confused with problem-based learning where
students are focused on resolving specific problems. Therefore, PBL is a
broader category of learning, where in addition to solving a specific
problem, students also address other areas that are not explicit in the
problem. From the lecturer’s point of view, PBL has authentic objectives
and uses real assessment, while the teacher acts as a coach with explicit
educational objectives [37]. On the other hand, from the student’s point of
view, PBL stimulates collaborative and cooperative learning, allows
continuous improvement in their presentations or products among others,
and is designed for the student to participate actively in the resolution of
the task [37]. Accordingly, the success of PBL is based on the design of
adequate problems to motivate self-study. In this context, RS and GIS are
areas of knowledge that allow the incorporation of new pedagogical
methodologies, such as, in this case, PBL [38–40].
Capítulo 2
pág. 82
3. UAV-RS in Agricultural Engineering at ETSIAM (University of
Cordoba)
The Higher Technical School of Agricultural Engineering (Escuela
Técnica Superior de Ingenieros Agrónomos, ETSIA) at the University of
Córdoba (Spain) began its academic activity in 1968, providing a degree in
agricultural engineering with various specialties. In 1989, it was authorized
to provide studies in forestry engineering. As a result, the school was called
the Higher Technical School of Agricultural and Forestry Engineering
(Escuela Técnica Superior Ingeniería Agronómica y de Montes, ETSIAM).
ETSIAM offers: three bachelor programs—Agrifood Engineering and Rural
Environment, Forestry Engineering, and Enology; two professional
master’s degrees—Agricultural Engineering, and Forestry Engineering;
and eight specialized master’s courses can be studied at ETSIAM. Since its
inception, ETSIAM has opted to incorporate the latest technological
advances in its academic programs. As an example, in 2018, a master’s
degree in digital transformation in the agri-food sector was implemented,
where the students learn aspects related to big data, cloud computing, IoT,
and others as applied to agriculture.
Since 2012, the Department of Graphic Engineering and Geomatics of
ETSIAM has established UAV-RS as one of its key areas of focus with
emphasis on research and teaching. Our own UAV-RS research allows our
department to consolidate areas of interest in agriculture which, thereby,
in terms of education, permits personal experience to form part of the
educational content. In addition, our research activity through projects
financed both publicly and privately, allows us to have the knowledge,
experience, and materials related to UAV technology, such as flight
platforms, sensors, and software. These three components are then made
available to students studying UAV-RS.
Three categories of students participate in UAV-RS programs:
bachelor students, master students, and PhD students. Bachelor students
do not study UAV-RS, however, in recent years, some have requested an
overview of these technologies, likely due to expectations and novelty. In
these cases, students do their final year project on RS-UAV in agriculture.
Master students have the opportunity to take a course on UAV-RS. PhD
students use UAV technology in their research activities.
Capítulo 2
pág. 83
RS is a discipline studied throughout the programs offered by
ETSIAM, while UAV studies are offered as an elective subject titled, “UAVs
in the Agroforest Sector” in the second course of the master’s degree in
agricultural engineering. This course has four ECTS (European Credit
Transfer and Accumulation System) with 40 hours of teacher-guided
activities and 60 hours of self-directed activities. The activity distribution is
summarized in Table 1. The purpose of this course is to provide students
an opportunity to work with UAS applied to precision agriculture using RS
techniques. Applying PBL, students process and analyze data that they
have registered themselves using RGB, multispectral, and thermal sensors
onboard UAV platforms. The main goal of this is for students to apply RS
methods in orthomosaics generated from images that have been registered
by sensors onboard UAV and to generate crop information to support
decision making.
Table 1. Type and duration of teacher-guided and self-directed
activities for "UAVs in the Agroforest Sector".
Teacher-guided
activities
Time duration
[hours]
Self-directed
activities
Time duration
[hours]
Master class 5 Analysis 20
Fieldwork 10 Study 10
Seminar 5 Group work 30
Group work 10
Team work 10
Figure 2 summarizes all the knowledge areas that students use
together with RS in this subject. In the Bachelor of Agrifood Engineering
and Rural Environment in ETSIAM, students have previously studied
geodesy, photogrammetry, GIS, and RS. Specifically, the courses in
photogrammetry and RS emphasize manned aerial and satellite platforms,
respectively and therefore, students have previously studied the
fundamentals of both geomatic disciplines. Through their academic
training, students have additionally studied mathematics, physics, and
statistics. Moreover, they have worked with R-commander developing
scripts as a statistical tool. Due to this preparation, students are ready to
work on the specific characteristics of the UAV-RS by using and applying
the knowledge previously acquired in earlier courses.
Capítulo 2
pág. 84
Figure 2. Knowledge areas involved in UAV-RS to transform a collection
of UAV-images into useful information.
4. Systems Design and Educational Activities
The application of UAV-RS in agriculture is relatively novel. UAV-RS
shares many technical and methodological aspects with satellite and
manned platforms. However, it is necessary to adapt to the specific
particularities that UAV-RS presents to the users. This explains the large
number of scientific publications in recent years, as shown in Figure 1. In
the “UAVs in the Agroforest Sector” course, students are required to
generate information about the state of a crop in terms of its vigor and
water needs, linking that information to the characteristics of the plot, such
as soil characteristics, topography or irrigation dose. It is, therefore, not
only a question of generating maps through RS but also of assessing the
“why” of the results obtained. For PBL to be successful, it is essential that
students collect their own data, which necessitates having the adequate
materials available. Figure 3 and Table 2 summarizes all the materials with
which the students work with: three flight platforms, four sensors to
register images, auxiliary materials for geometric and atmospheric
corrections, as well as different software solutions for the treatment of
images. In addition, ETSIAM manages a “nature classroom” with different
crops. Among them, there is an orchard with an area equal to 10 hectares
of high-density hedgerow olive trees (north latitude 37°56′05′′, west
longitude 4°42′59′′, WGS 84) (Figure 4) with four different variables to be
analyzed: phenotype, density, orientation, and irrigation dose together
with soil characteristics. This provides students a real-world case study
where they can collect data via RS techniques and then convert them into
information useful for assisting decision-making. This “nature classroom”
Capítulo 2
pág. 85
or plot, then, is the central element on which the PBL methodology is
designed.
Figure 3. Materials provided to students: (a) UAV platforms, (b) sensors,
(c) auxiliary materials and (d) software.
Capítulo 2
pág. 86
Table 2. Technical characteristics of the UAV platforms, sensors,
auxiliary materials, and software used by students.
Material Description
UAV Platforms
Multirotor
MD4-1000
Quadcopter with a maximum payload equal to 1.2
kg. Maximum flight time 30 minutes.
DJI Mavic
Pro2
Quadcopter with a Hasselblad RGB sensor.
Maximum flight time 30 minutes.
Elimco E-300 Fixed wind UAV with a maximum payload equal
to 4 kg. Maximum flight time 90 minutes.
Sensors
Sony Nex7 RGB sensor.
Tetracam
Mini MCA-6
Multispectral sensor with six individual sensors,
one for each band, arranged in a 2 × 3 array. Spectral
bands equal to 450 nm, 530 nm, 670 nm, 700 nm, 740
nm, and 780 nm.
Xeneth Gobi
640
Uncooled thermal sensor. It delivers raw digital
images at 16 bits. Dynamic range from –20 ºC to 120
ºC, spectral resolution equal to 0.05 ºC.
Parrot
Sequoia
Multispectral sensor with an RGB sensor and four
individual spectral bands at 550 nm, 660 nm, 735
nm, and 790 nm, and a sunshine sensor.
Auxiliary
material
ASD
HandHeld-2
Portable spectroradiometer. Spectral range from
325 nm to 1075 nm.
Calibration
panels 0.5 × 0.5 m reflectance targets.
Flir E60 Heat gun. Dynamic range from –20 ºC to 120 ºC.
Resolution equal to 0.05 ºC
Leica GS15 GNSS multi-frequency receiver. Horizontal
accuracy 8 mm +0.5 ppm at real-time kinematic
measurement.
Vantage Pro2 Wireless weather station to measure temperature
and relative humidity of the air, direction, and
speed of the wind and atmospheric pressure.
Software
Pix4D Photogrammetry software suite for UAV mapping.
R Software for statistical computing and computing.
QGIS Desktop Geographic Information System
Capítulo 2
pág. 87
Figure 4. Location of the “nature classroom”.
While students learn UAV RS, PBL allows them to acquire other
competencies grouped into three classes: basic, general, and specific. Basic
competencies are those referring to transversal competencies, transferable
to many functions and tasks (communication, teamwork, etc.). While
general competencies are those referring to successful integration into
social and working life (language, new technologies, etc.). Finally, specific
competencies are those directly related to a particular area of knowledge.
Regarding the first group, each student learns in a highly personalized and
independent way within their development and often close to the context
of the research. In addition, they solve problems in new or unfamiliar
environments related to their area of study, such as programming.
Likewise, they acquire skills that will allow them to continue studying
autonomously. For general competencies, students develop the ability to
apply the acquired knowledge to solve problems, analyze the information,
and synthesize it to facilitate the decision-making process. In addition, the
ability to transmit information and conclusions using new communication
channels is developed. Finally, and with specific character, students
acquire knowledge and skills to be applied in the agricultural sector.
Table 3 summarizes teacher-guided activities and Figure 5 shows the
temporal distribution. Lectures have a total duration of five hours. Firstly,
students are given the rationale for learning these technologies and their
applications to precision agriculture, advantages and disadvantages in
terms of spatial and temporal resolution compared to other platforms, as
well as when their use is justified (Table 3, L1). In addition, the student
learns the legislative aspects that regulate the use of this type of platform,
highlighting that today, the current legislation only allows the use of RPAS.
Once the RPAS concept is introduced, the student learns what a UAS is and
Capítulo 2
pág. 88
the different sub-systems that it is composed of: the flight platform, sensors,
and the control station (Table 3, L2). Flight platforms are analyzed from
different points of view, such as architecture, autonomy, weight, and
maximum flight height, providing an overview of their individual
capabilities, features, and versatilities. In another session, the student
learns the different sensors of interest in agriculture to be onboard a UAV
(Table 3, L3): RGB, multi/hyper-spectral, thermographic, and LiDAR. An
overview of the current state of sensors is made, assessing development
level, possibilities, and opportunities for improvement. Finally, concepts in
photogrammetry are reviewed as this was studied in the bachelor’s degree
(Table 3, L4) and only needs to be extended conceptually to unmanned
platforms. Overall, depending on the type of sensor onboard the UAV and
our experience with it, particularities and suggestions are detailed for the
students.
Three seminars based on UAV-RS applications in agriculture using
RGB, multispectral and thermal sensors, and one on diffusing information
form part of the course. Based on scientific articles, web pages, and our own
results from UAV flights, students learn different indexes and their
usefulness in agriculture, such as color indices, isolating vegetation based
on RGB sensors, determining the water needs through vegetation,
temperature indices and so forth. These seminars are combined over time
with fieldwork as the different applications of the sensor type are studied,
thereby allowing students to put into practice what they have learned in
the classroom. Each fieldwork assignment starts with a UAV flight. Thus,
students plan the flight, use GNSS receivers to measure ground control
points for geometric correction, use a thermographic gun and
spectroradiometer to measure calibration panels, and then perform an
atmospheric correction using the empirical line method.
The transformation of the UAV collected data to useful information
for the farmer implies the generation of a series of products, such as maps
or graphs, that have to be clear and easy to interpret. Today, aside from the
classic analogical maps, it is necessary to provide this information digitally.
For this reason, the last seminar (Table 3, S4) has been designed to present
different tools for publishing geographic information through web services
and applications.
Finally, three workshops are organized throughout the course. The
first (Table 3, W1) is focused on the generation of digital surface models
(DSM), digital elevation models (DEM) and orthomosaics from an image
collection obtained from a UAV flight. The second (Table 3, W2) focuses on
statistical treatments of images, atmospheric correction, and generation of
Capítulo 2
pág. 89
indices using R-commander. In the last workshop (Table 3, W3), students
learn to automate spatial analysis processing using QGIS and the Python
programming language.
Table 3. Type and duration of teacher-guided activities for
"UAVs in the Agroforest Sector".
Time duration
[hours]
Lectures
L1: Introduction and legal regulation. 1
L2: UAS and UAVs 1
L3: Sensors 1
L4: Flight planning and processing a UAV
flight 2
Seminars
S1: RGB sensor applications 1
S2: Multispectral sensor applications 1,5
S3: Thermal sensor applications 1,5
S4: Diffusion 1
Fieldwork
F1: UAV flight RGB sensor 3
F2: UAV flight multispectral sensor 3,5
F3: UAV flight thermal sensor 3,5
Workshop
W1: UAV flight processing 2
W2: Manage raster images using R-
commander 4
W3: Process automation using QGIS 4
Figure 5. Temporal distribution of teacher-guided activities, lectures,
seminars, fieldwork, and workshops for "UAVs in the Agroforest sector".
Capítulo 2
pág. 90
Teacher-guided activities support self-directed activities with the
main objective being that the students are able to independently provide
information about crops. Table 4 summarizes self-directed activities
according to the sensor used. All in-course UAV flights have a common
goal, to generate a DSM, a DEM, and an orthomosaic. Upon completion of
the UAV flight, the students measure auxiliary data to perform geometric
corrections from the measurements of the ground control points (GCPs)
obtained from the GNSS receivers and atmospheric corrections obtained
from the spectroradiometer or thermographic gun. In addition,
environmental parameters such as air temperature, relative humidity, and
atmospheric pressure are measured for each UAV flight. A true- color
ortho-mosaic is produced using RGB images from the RGB UAV flight (F1),
which is used to isolate hedgerow olive trees by a color filter. Scientific
references, such as [41], are provided to the students so that they may
reproduce the processing by developing an R-Commander script. Different
color filters, like ExG or ExR, are implemented to later assess which one
provides the best results in terms of isolating vegetation from the soil. In
addition, these results are applied again in the multispectral and
thermographic orthomosaics to extract the information only from the crop.
For the multispectral UAV flight (F2), once the spectral orthomosaic
for each spectral band is generated, the students apply an atmospheric
correction using the empirical line method. An R-Commander script is
developed by the students to calculate a lineal model to relate image values
and spectroradiometer measurements for each spectral band, which is later
applied to generate a new set of atmospheric-corrected orthomosaics.
Using these atmospheric-corrected orthomosaics, students have to
calculate five vegetation indexes (VIs). NDVI and SAVI are obligatory
while the students are free to calculate the rest. The calculation of these VIs
is done through a script developed in Python and QGIS. Subsequently, the
value of the VI relative to the crop is isolated using the mask generated
from the RGB UAV flight in the previous activity and each hedgerow is
statistically characterized. In addition, every tree is analyzed individually
since the plantation pattern is known. This is carried out for each VI
calculated so that the students once again have to automate the whole
process through Python and QGIS.
Finally, a thermal orthomosaic is generated using images from a
thermographic UAV flight (F3). After, atmospheric correction is applied
through the empirical line method using in-field temperature
measurements of panel calibrations. For that, students use the R-
Commander script developed previously. Then, every hedgerow and the
individual trees are thermally characterized using basic statistics.
Capítulo 2
pág. 91
Moreover, a crop water stress index (CWSI) model from scientific literature
is applied [42].
As a result of the application of RS techniques, the student
characterizes the crop, both at the level of hedgerow and individual trees.
It is then necessary for the student to learn tools that agilely show the
results obtained. As the final step, currently, the student learns Carto
(www.carto.com) as a cloud-based GIS tool to present their results on
different devices (tablets, smartphones or laptops). With all the information
collected from the crop, the student analyzes the causes that explain the
differential behavior in terms of vigor, temperature, and water stress, and
generate a final report with all the results obtained.
Table 4. Input data and results of each of the self-directed
activities carried out by students linked to fieldwork and
seminars.
Teacher-guided
activity
Input data Results
F1: RGB UAV
flight
RGB images
GCPs
Orthomosaic,
DSM and DEM
Vegetation masks
Analysis and
interpretation
F2: Multispectral
UAV flight
MSI images
GCPs
Spectroradiometer
measurements
Spectral-
orthomosaic
Atmospheric
correction
Vegetation
indexes
processing
Crop line
characterization
Single tree
characterization
Analysis and
interpretation
F3:
Thermographic
UAV flight
Thermal images
GCPs
Heat gun
measurements
Weather station
measurement
Thermal-
orthomosaic
Atmospheric
correction
Crop Water Stress
Index
Capítulo 2
pág. 92
Crop line
characterization
Single tree
characterization
Analysis and
interpretation
S4: Diffusion Orthomosaics from
F1, F2 and F3.
Geographic data
viewer web
5. Results
Students participate actively in the learning process by collecting field
data, developing scripts, and analyzing results (Figure 6). As such, from the
beginning the students are motivated and involved in the learning process,
awakening the desire to learn new tools, methodologies, and RS methods
with direct application in agriculture. This mainly stems from two aspects.
Firstly, students have access to and use of sensors, flight platforms, and
scientific-professional instruments acquired by the research group over the
years. Secondly, being able to do real fieldwork on an experimental farm
allows the results from RS to be subsequently validated.
Capítulo 2
pág. 93
Figure 6. Some teacher-guided activities: (a) learning about UAV
platforms and (b) fieldwork to measure temperature and spectral
signature for atmospheric corrections.
As a result of the learning process, some of the products produced by
students are presented as examples in Figure 7. Using multispectral
sensors, students calculate different vegetation indices (Figure 7a). Once
these differences associated with the vigor of the tree have been detected,
the student must analyze the cause. In the same manner, the students
analyze the results obtained in the thermographic orthomosaic (Figure 7b).
As in the previous case, they validate the results obtained in the field. Being
able to work in the "nature classroom" allows students to personally check
the characteristics of the soil or the development of the crop. In the case of
Figure 7b, a relationship is shown between the temperature map and the
photographs of the crop that explain and justify these differences. Finally,
Figure 7c shows part of a water stress map. In this case, the students found
out which area of the plot was under deficit irrigation.
Capítulo 2
pág. 94
Figure 7. Examples of student results: (a) comparison of results of
different vegetation indices, (b) analysis of the relationship between
thermographic UAV and field inspection, and (c) a water stress map.
To date (2018/19 academic year), the course “UAVs in the Agroforest
Sector” has been taught over four academic years. In these four editions,
the University of Córdoba has surveyed students through the teaching
quality system, currently with full data from the first three courses. Table 5
summarizes a total of 21 questions evaluated between 0 (lowest
qualification) and 5 (highest qualification) represented in three sections
including course planning, course development, and learning evaluation.
Since the course’s inception, students have evaluated the subject positively,
marking high qualifications, close to 5 in all sections. Regarding the
Capítulo 2
pág. 95
“Course Planning“ section, students know from the first day of the course
how the subject is planned, what they are intended to do, and what
capacities are to be developed and achieved throughout the course.
As related to the “Course Development” section, students evaluate
whether adequate resources have been used, if the bibliography and other
sources of information are useful, if the lecturer explains the contents with
clarity, and if the students are interested in the subject. This section, having
scored 4.56 out of 5 in the first year, is the one that has improved the most.
This improvement is the sum of several factors. Firstly, the research of the
teaching staff in UAV-RS is directly projected in the quality of the subject.
In addition, having access to adequate instruments is an added benefit that
students value positively.
In the “Learning Evaluation” section, students answer two questions:
a) Do I know what will be required of me to pass the subject? and b) Are
the criteria and evaluation systems adequate? The feedback received from
the students encourages the department to continue with this teaching
system. Year by year, students score this section higher demonstrating that
students favor PBL to traditional methods, which in turn, leads to greater
student engagement.
Table 5. Results of the student satisfaction survey (Values in the
range between 0 and 5).
Section 2015/16 2016/17 2017/18
Course Planning 4,25 5,00 4,83
Course Development 4,56 4,69 4,77
Learning Evaluation 4,67 4,71 4,81
In addition to the results of the quality survey, our personal
impression is that students value this teaching method. Being able to use
these technologies in a real world-simulation field they can visit whenever
they want makes them feel involved in the learning process, which results
in stronger educational drivers. Of all the content of the subjects, the
development of scripts is the most difficult task for them. Even so, each
year students are more receptive to this kind of challenge, likely because
they are progressively more aware of the undergoing technological
changes in agriculture.
Although the results are positive, in order to improve the quality of
the subject it is important to keep the subject dynamic, adapted, and
updated in content. For this reason, the research group will start working
with hyperspectral sensors in 2019. Once the knowledge and experience
Capítulo 2
pág. 96
acquired are adequate, these sensors will be introduced into the learning
process. Furthermore, additional crops will be introduced and students
will be distributed into workgroups. Each group will center on an assigned
crop and the results obtained will be shared among the groups.
6. Conclusions
The university student has to acquire knowledge and skills adapted to
what the labor market demands and, therefore, academic content must
meet these standards. In this way, UAVs have to be introduced in higher
education to teach future professional skills related to agricultural
practices. This manuscript has shown how ETSIAM at the University of
Cordoba has introduced UAV-RS in its educational activities using PBL.
The results from the quality surveys conducted on our students have
shown a very high degree of satisfaction. In our opinion this stems from the
research activities of the professors and the selected teaching method. The
research activity allowed our educators to gain experience and knowledge
on UAV-RS and the possibility to acquire materials through privately and
publicly financed projects. Moreover, PBL has shown to be conducive to
the learning process. In this context, students are motivated to learn and
engage in the subject from the beginning, as they see the link between the
subject, their own professional interest, and what companies demand.
Active teaching methods, such as flipped classroom or team-based
learning, are increasingly exposing students to new educational models,
which are designed to aid students in fully understanding course material
and their applications. These methods put greater emphasis on student
learning and it gives them greater impetus in the process of learning.
Therefore, the roles of students are changing from passive to active
participants. What this means is that educators are constantly
experimenting with teaching strategies and as such need to have platforms
to share successes and failures in order to cultivate a more productive
learning culture.
Author Contributions: Conceptualization, F-J.M-C. and A.G-F.;
methodology, F-J.M-C., F.P.P., P.T-T., J.E.M.L., and A.G-F.; resources,
F-J.M-C. and A.G-F.; writing—original draft preparation, F-J.M-C.;
writing—review and editing, F.P.P. and P.T-T.
Funding:
Acknowledgments: The authors thank the support of the Higher
Technical School of Agricultural and Forestry Engineering (Escuela
Capítulo 2
pág. 97
Técnica Superior Ingeniería Agronómica y de Montes) of the
University of Córdoba, Spain.
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Foresman, T.W.; Cary, T.; Shupin, T.; Eastman, R.; Estes, J.E.; Faust, N.; Jensen,
J.R.; Kemp, K.K. Internet teaching foundation for the Remote Sensing Core
Curriculum program. ISPRS J. Photogramm. Remote Sens. 1997, 52, 294–300,
doi:10.1016/S0924-2716(97)00025-7.
2. Barkley, E.F. Student Engagement Techniques: A Handbook for College
Faculty; John Wiley & Sons: 2009.
3. Coates, H. Student Engagement in Campus-Based and Online Education:
University Connections; Routledge: 2006.
4. Romero, R.; Ferrer, A.; Capilla, C.; Zunica, L.; Balasch, S.; Serra, V.; Alcover, R.
Teaching Statistics to Engineers: An Innovative Pedagogical Experience. J.
Stat. Educ. 1995, 3, doi:10.1080/10691898.1995.11910481.
5. Behar Gutiérrez, R.; Grima Cintas, P. La Estadística en la Educación Superior
¿Formamos Pensamiento Estadístico? Ing. Compet. 2011, 5, 84–90,
doi:10.25100/iyc.v5i2.2299.
6. Union, E. Copernicus. Europe’s eyes on Earth. Available online:
https://www.copernicus.eu/ (accessed on 15 April 2019).
7. Gorelick, N.; Hancher, M.; Dixon, M.; Ilyushchenko, S.; Thau, D.; Moore, R.
Google Earth Engine: Planetary-scale geospatial analysis for everyone.
Remote Sens. Environ. 2017, 202, 18–27, doi:10.1016/j.rse.2017.06.031.
8. Council, N.R. Future US Workforce for Geospatial Intelligence; National
Academies Press: 2013.
9. United States Department of Labor. High Growth Industry Profile—
Geospatial Technology; 2010.
10. Gewin, V. Mapping opportunities. Nature 2004, 427, 376–377,
doi:10.1038/nj6972-376a.
11. Jensen, J.R.; Dahlberg, R.E. Status and content of remote sensing education in
the United States. Int. J. Remote Sens. 1983, 4, 235–245,
doi:10.1080/01431168308948543.
12. Dell'Acqua, F.; Pasca, L. Technical Education in the European University
System on Aerospace and Remote Sensing: A Year 2013 Review [Education].
IEEE Geosci. Remote Sens. Mag. 2014, 2, 29–33,
doi:10.1109/MGRS.2014.2304131.
13. Fras, M.K.; Grigillo, D. Implementation of active teaching methods and
emerging topics in photogrammetry and remote sensing subjects. Int. Arch.
Photogramm. Remote Sens. Spat. Inf. Sci. 2016, 41, doi:10.5194/isprsarchives-
XLI-B6-87-2016
Capítulo 2
pág. 98
14. Azzalis, L.; Sato, S.; de Mattos, M.; Fonseca, F.; Giavarotti, L. Active learning
versus traditional teaching. Rev. Ensino Bioquím. 2009, 7, 2.
15. Weltman, D. A Comparison of Traditional and Active Learning Methods: An
Empirical Investigation Utilizing a Linear Mixed Model; University of Texas:
Austin, TX, USA, 2008.
16. Bonwell, C.C.; Eison, J.A. Active Learning: Creating Excitement in the
Classroom; ERIC Digest: 1991.
17. Guida, R. Introduction of podcasts in remote sensing education. In
Proceedings of the 2010 IEEE International Geoscience and Remote Sensing
Symposium, Honolulu, HI, USA, 25–30 July 2010; pp. 1104–1106.
18. Mountrakis, G.; Triantakonstantis, D. Inquiry-Based Learning in Remote
Sensing: A Space Balloon Educational Experiment. J. Geogr. High. Educ. 2012,
36, 385–401, doi:10.1080/03098265.2011.638707.
19. Baldina, E.A.; Chalova, E.R.; Knizhnikov, Y.F.; Tutubalina, O.V. Remote
sensing education using Internet—Prospects of the Inter-University
Aerospace Centre. In Proceedings of the IEEE International Geoscience and
Remote Sensing Symposium, Toronto, Canada, 24–28 June 2002.
20. Joyce, K.E.; Boitshwarelo, B.; Phinn, S.R.; Hill, G.J.E.; Kelly, G.D. Interactive
online tools for enhancing student learning experiences in remote sensing. J.
Geogr. High. Educ. 2014, 38, 431–439, doi:10.1080/03098265.2014.933404.
21. Colomina, I.; Molina, P. Unmanned aerial systems for photogrammetry and
remote sensing: A review. ISPRS J. Photogramm. Remote Sens. 2014, 92, 79–
97, doi:10.1016/j.isprsjprs.2014.02.013.
22. Singh, K.K.; Frazier, A.E. A meta-analysis and review of unmanned aircraft
system (UAS) imagery for terrestrial applications. Int. J. Remote Sens. 2018, 39,
5078–5098, doi:10.1080/01431161.2017.1420941.
23. Mazur, M.; Wisniewski, A.; McMillan, J. Clarity from Above. PwC Global
Report on the Commercial Application of Drone Technology; 2016.
24. Maes, W.H.; Steppe, K. Perspectives for Remote Sensing with Unmanned
Aerial Vehicles in Precision Agriculture. Trends Plant Sci. 2019, 24, 152–164,
doi:10.1016/j.tplants.2018.11.007.
25. Al-Tahir, R. Integrating UAV into geomatics curriculum. Int. Arch.
Photogramm. Remote Sens. Spat. Inf. Sci. 2015, 40, 387.
26. Elaksher, A. Modernizing the Photogrammetry Curricula with Small
Unmanned Aerial Systems. Surv. Land Inf. Sci. 2018, 77, 75–84.
27. Pereira, E.R.; Zhou, S.; Gheisari, M. Integrating the use of UAVs and
photogrammetry into a construction management course: Lessons learned. In
the International Symposium on Automation and Robotics in Construction;
IAARC Publications: 2018; pp. 1–8.
28. Matthew, R.G.S.; Hughes, D.C. Getting at deep learning: A problem-based
approach. Eng. Sci. Educ. J. 1994, 3, 234–240.
Capítulo 2
pág. 99
29. Lattimer, H.; Riordan, R. Project-Based Learning Engages Students in
Meaningful Work. Middle Sch. J. 2011, 43, 18–23,
doi:10.1080/00940771.2011.11461797.
30. Wurdinger, S.; Haar, J.; Hugg, R.; Bezon, J. A qualitative study using project-
based learning in a mainstream middle school. Improving Sch. 2007, 10, 150–
161, doi:10.1177/1365480207078048.
31. Delisle, R. How to Use Problem-Based Learning in the Classroom; 1997.
32. Brodeur, D.R.; Young, P.W.; Blair, K.B. Problem-based learning in aerospace
engineering education. In Proceedings of the 2002 American Society for
Engineering Education Annual Conference and Exposition, Montreal,
Canada.
33. Gijselaers, W.H. Connecting problem-based practices with educational theory.
New Dir. Teach. Learn. 1996, 1996, 13–21, doi:10.1002/tl.37219966805.
34. Martinez, M.E.J.P.d.k. What is metacognition? 2006, 87, 696–699.
35. Barrows, H.S. Problem-based learning in medicine and beyond: A brief
overview. 1996, 68, 3–12, doi:10.1002/tl.37219966804.
36. Harwell, S. Project-Based Learning. Promising Practices for Connecting High
School to the Real World; 1997; Volume 2328.
37. Martí, J.A.; Heydrich, M.; Rojas, M.; Hernández, A. Aprendizaje basado en
proyectos. Rev. Univ. EAFIT 2010, 46.
38. Drennon, C. Teaching Geographic Information Systems in a Problem-Based
Learning Environment. J. Geogr. High. Educ. 2005, 29, 385–402,
doi:10.1080/03098260500290934.
39. Botti, J.A.; Myers, R. Exploring the environment: A problem-based approach
to learning about global change. In Proceedings of the 1995 International
Geoscience and Remote Sensing Symposium, IGARSS’95. Quantitative
Remote Sensing for Science and Applications, Firenze, Italy, 10–14 July 1995.
40. Croft, S.K.; Myers, R.J. Helping students and teachers make sense of remote
sensing via the Internet. In Proceedings of the IGARSS’96. 1996 International
Geoscience and Remote Sensing Symposium, Lincoln, NE, USA, 31 May 1996.
41. Guijarro, M.; Pajares, G.; Riomoros, I.; Herrera, P.J.; Burgos-Artizzu, X.P.;
Ribeiro, A. Automatic segmentation of relevant textures in agricultural
images. Comput. Electron. Agric. 2011, 75, 75–83,
doi:10.1016/j.compag.2010.09.013.
42. Egea, G.; Padilla-Díaz, C.M.; Martinez-Guanter, J.; Fernández, J.E.; Pérez-Ruiz,
M. Assessing a crop water stress index derived from aerial thermal imaging
and infrared thermometry in super-high density olive orchards. Agric. Water
Manag. 2017, 187, 210–221, doi:10.1016/j.agwat.2017.03.030.
© 2019 by the authors. Licensee MDPI, Basel,
Switzerland. This article is an open access article
distributed under the terms and conditions of the
Capítulo 2
pág. 100
Creative Commons Attribution (CC BY) license
(http://creativecommons.org/licenses/by/4.0/).
Capítulo 3
pág. 101
6 CAPÍTULO 3
Publicado en: Remote Sensing, 2020, 12(14), 2210
https://doi.org/10.3390/rs12142210
Article
Effect of Lockdown Measures on
Atmospheric Nitrogen Dioxide During
SARS-CoV-2 in Spain
Francisco-Javier Mesas-Carrascosa *, Fernando Pérez Porras, Paula
Triviño-Tarradas, Alfonso García-Ferrer and Jose Emilio Meroño-
Larriva
Department of Graphic Engineering and Geomatics, Campus de Rabanales,
University of Cordoba, 14071 Córdoba, Spain; [email protected] (F.P.P.);
[email protected] (P.T.-T.); [email protected] (A.G.-F.); [email protected] (J.E.M.-L.)
* Correspondence: [email protected]
Received: 01 June 2020; Accepted: 08 July 2020; Published: date
Abstract: The disease caused by SARS-CoV-2 has affected many countries
and regions. In order to contain the spread of infection, many countries
have adopted lockdown measures. As a result, SARS-CoV-2 has
negatively influenced economies on a global scale and has caused a
significant impact on the environment. In this study, changes in the
concentration of the pollutant Nitrogen Dioxide (NO2) within the
lockdown period were examined as well as how these changes relate to
the Spanish population. NO2 is one of the reactive nitrogen oxides gases
resulting from both anthropogenic and natural processes. One major
source in urban areas is the combustion of fossil fuels from vehicles and
industrial plants, both of which significantly contribute to air pollution.
The long-term exposure to NO2 can also cause severe health problems.
Remote sensing is a useful tool to analyze spatial variability of air quality.
For this purpose, Sentinel-5P images registered from January to April of
2019 and 2020 were used to analyze spatial distribution of NO2 and its
evolution under the lockdown measures in Spain. The results indicate a
significant correlation between the population’s activity level and the
reduction of NO2 values.
Keywords: nitrogen dioxide; SARS-Cov-2; Sentinel-5P; air pollution
Capítulo 3
pág. 104
1. Introduction
Clean air is an essential requirement for human health, and as such air
pollution is a major threat to human well-being. Air pollution is the largest
environmental health risk in many regions around the world. The World
Health Organization estimates that air pollution kills 7 million people
worldwide every year, making it necessary to monitor air pollution and
improve air quality [1]. The environmental impact is more evident in areas
where population density is high, being particularly severe in megacities
where high population density, extensive motor vehicle use and strong
industrial expansion are combined [2]. Poor air quality is not exclusive to
megacities; even small cities with populations around 150,000 can have this
problem [3]. Therefore, the economic development of cities with expanding
industrial areas is associated with increasing population size and
environmental degradation of the surrounding areas [4]. In addition, the
high levels of motor vehicle activity [5] and their inappropriate use [6]
cause an increase of air pollutants in city centers [7]. As a result, quality of
life and human health is worsening specifically in cardiovascular,
neurological, and respiratory diseases [8–10] and even results in higher
mortality rates [11,12]. As a consequence of this, developed and developing
countries are more and more attentive to urban air quality, developing
guidelines, directives and standards to inform and support policymakers
[13,14] to reduce the health impacts of air pollution.
Common pollutants in the troposphere, the innermost layer of Earth’s
atmosphere, include ozone (O3), carbon monoxide (CO), sulfur dioxide
(SO2), nitrogen dioxide (NO2) and aerosols. NO2, specifically, has been
correlated with mortality in studies in different parts of the world [15,16].
It is true, however, that there is no clear evidence to establish that NO2 acts
as an independent agent causing increases in the mortality rate [17]. Rather,
it is widely believed that NO2 could act as a substitute component for others
that are not currently being monitored or, more broadly, as a mixture of
pollutants [18]. The result is that NO2 is included in the multi-pollutant
health indexes [19]. Epidemiological research and studies have shown how
NO2 is related to adverse health effects like lung cancer [20,21], asthma
exacerbations [22,23] and cardiopulmonary mortality [24,25]. Mainly, NO2
forms from ground-level emissions caused by the burning of fossil fuels
from industrial sources, vehicles and power plants. It contributes to
ground-level ozone formation and it is linked to negative effects on
respiration. For example, NOx reacts with moisture, ammonia and other
compounds to form small particles that can penetrate into sensitive parts
of the lungs.
Capítulo 3
pág. 105
Traditionally, air pollution is monitored using a networks of sensors,
such as gas chromatograph-mass spectrometers [26] or ultraviolet sensors
[27], among others, which are distributed over a territory and provide
quality information on a wide range of pollutants. The traditional
instrumentation used for air quality monitoring is expensive, large,
location dependent and yields extremely low spatial and temporal
resolution [28]. For this reason, portable environmental sensor systems
have been developed using Wireless Sensor Network technology at a lower
cost, offering data with a higher frequency over time. They are also easier
to relocate and provide better coverage of the area of interest due to
allowing the use of a larger number of nodes [29,30], permitting the
development of more efficient and accurate air quality models [31]. Despite
these advantages, however, it is not possible to map a broad region.
Advances in atmosphere remote sensing have opened new avenues for
measuring and monitoring atmospheric pollution at local, regional,
continental or global scales [32], providing new challenges and
opportunities for environmental health research [33]. The ability to observe
and monitor air pollutants from sensors onboard satellite platforms has
improved in the last two decades. From the first ultraviolet-visible
spectrometer, the Global Ozone Monitor Experiment (GOME), with a
spatial resolution equal to 40 × 320 km2 [34], followed by the SCanning
Imaging Absorption SpectroMeter for Atmospheric CHartography
(SCIAMACHY), with a pixel size of 30 × 60 km2 [35], and GOME-2, 40 × 80
km2 [36] to the Ozone Monitoring Instrument, 13 × 24 km2 [37] it has been
possible to study the distribution of pollutants at urban scales. Recently,
the European Space Agency launched the Sentinel-5 Precursor (S5P) to
provide data on air quality, the climate and the ozone layer using the
TROPOspheric Monitoring Instrument (TROPOMI) as its payload [38]
providing a significant improvement in data quality and spatial resolution
now at 7 × 7 km2 [39]. The spectral bands of the spectrometer TROPOMI
ranges from ultraviolet, visible, near infrared and shortwave infrared,
allowing the observation of the prevalence of aerosols in the atmosphere,
cloud characteristics, concentrations of carbon monoxide (CO),
formaldehyde (CH2O), nitrogen dioxide (NO2), ozone (O3), sulphur dioxide
(SO2) and methane (CH4). The Tropospheric Vertical Column Density
(VCD) data of these components are measured from space by sensors like
TROPOMI which serves as an accurate proxy at ground level in many air
quality applications [40]. The VCD is defined as the number of molecules
of a certain atmospheric gas between the on-board sensor of the satellite
platform and the Earth’s surface per unit area. Tropospheric and
stratospheric column densities are separated using a data assimilation
system based on the three-dimensional global Tracer chemical Transport
Capítulo 3
pág. 106
Model (TM5-MP), after which they are converted to VCD by a look-up table
of altitude dependent air-mass factors and information on the vertical
distribution of NO2 [41]. For NOx, VCD measurements have been
successfully used to estimate trends and variations in atmospheric
concentration [42,43], infer surface emissions [44,45] and monitor emission
changes at a given location [46,47].
Since mid-February 2020, all efforts of many countries were directed
towards combating SARS-CoV-2. However, at the beginning of 2020, the
risk of a pandemic from a virus was not among the perceived risks
worldwide. This year was the first time that the World Economic Forum’s
Global Risk Report showed how climate change and environmental risks
were among the top positions [48]. However, both problems are associated,
the origin of new pathogens, such as SARS-CoV-2, may be explained by
environmental degradation. The coronaviruses have been known about
since 1930 [49]. They are transmitted from animals [50] and have been
increasing in number over the last decades [51]. The degradation of natural
spaces from human activity is increasing the rate of contact between wild
spaces and humans, resulting in new diseases and facilitating their
expansion [52–54]. Nevertheless, while many believe that the climate, and
therefore the environment is changing, some think this is not attributable
to human activity [55], which may be due to their perception of cultural
values [56] or ideologically and politically motivated actors [57].
Unfortunately, the spread of the coronavirus SARS-Cov-2 has been
unstoppable and has become a pandemic [58], with dramatic results in
countries like Spain, Italy and the United Kingdom [59] in Europe’s case.
Interventions like quarantine or isolation have shown to be effective in
reducing the number of SARS-CoV-2 infections [60]. In addition, some
countries and regions have deemed it necessary to impose lockdown
measures on economic activities and with unprecedented travel restrictions
[61,62]. Under this lockdown period, changes in air pollution can provide
valuable information on air quality improvement when there are
restrictions on emissions. In this manuscript, using Sentinel S5P images, we
analyze the spatial and temporal variation of NO2 concentrations during
Spain’s SARS-Cov-2 lockdown phase which took place in March and April
2020, and how these variations are related to city-scale demographics.
2. Materials and Methods
2.1. Study Area
Capítulo 3
pág. 107
Spain has just over 47 million inhabitants as of 1 January 2020.
Although unevenly distributed throughout the territory, the average
population density is equal to 82 habitants per km2. Figure 1 shows
population density using the European Environment Agency 10 × 10 km
grid as a reference. There are areas with highly concentrated populations
next to areas of demographic voids. The factors that explain this
unbalanced population distribution in Spain are both natural and
historical. Regarding natural factors areas with flat and low-lying relief,
with a temperate and humid climate and access to the sea and rivers are the
most populated. As far as historical factors are concerned, the distribution
of the population is related to the economic structure of the country and
the development of transport and communication infrastructure. As a
result, the average population density in Spain is 416 inhabitants per square
kilometer, with very high density areas, over 750 habitants per square
kilometer, compared to others with very low densities.
Figure 1. Distribution of population density in Spain on a 10 × 10 km
grid.
Figure 2 shows the location of cities with more than 275,000
inhabitants, with the highest concentrations in Madrid, having more than
3 million inhabitants, and Barcelona, with just over 1.6 million. The
relationship between the number of inhabitants in these eleven cities and
the variation of NO2 VCD under the lockdown measures is the focus of this
analysis.
Capítulo 3
pág. 108
Figure 2. Location of the eleven most populated cities in Spain.
2.2. Remote Sensing Image Collections
The TROPOMI on-board the Sentinel-5 Precursor (Sentinel-5P) was
used to collect data on NO2 concentration. The Sentinel-5P mission,
launched by the European Space Agency in 2017, is a low-orbit polar
satellite used to monitor Earth’s atmosphere with a high spatio-temporal
resolution using the TROPOMI. Concretely, it is a multispectral sensor that
registers reflectance values at ultraviolet-visible (250–500 nm), near-
infrared (675–775 nm) and short-wave infrared (2305–2385 nm)
wavelengths which measures concentrations of ozone, methane,
formaldehyde, aerosols, carbon monoxide, nitrogen oxide and sulphur
dioxide as well as cloud characteristics like cloud fraction, cloud base and
pressure.
Image processing was performed with the Google Earth Engine, a
cloud-based platform for geospatial analysis with high computational
capabilities [63]. A total of 1637 Sentinel 5P Nitrogen Dioxide level-3 scenes
of Spain from January to April of 2019 were used and 1636 scenes from the
same period were used from 2020 (Table 1). Thus, Sentinel 5P Level-2 data
[64] are processed to obtain a single grid per orbit, which allows the Google
Earth Engine to process the data. In addition, the data are previously
filtered, resulting in pixels with quality assurance values less than 75%
being removed, such as cloud or partially snow-covered pixels, errors and
or problematic retrievals. First, for each month and year a median image
was generated to represent the time series of the images, obtaining an
individual image for each time period. Thus, each pixel in the output image
Capítulo 3
pág. 109
was equal to the median value of all the images at that location. On the NO2
median images, once masked for the geographical space of Spain, the
statistics corresponding to maximum, minimum, median, and 2nd and 3rd
quartile were determined. This allowed a space-time comparison of the
NO2 VCD throughout the year and between years.
In addition, the variation of NO2 VCD before and after the lockdown
measures was determined. Thus, taking 15 March 2020 as a time reference,
the day on which the lockdown measures became effective in Spain, a
median image was calculated to represent the NO2 VCD one month before
and after the adoption of these measures, and the variation of this
component was then determined. For the most populated cities in Spain,
the average value of variation of NO2 VCD was determined in order to
analyze its relationship with the number of inhabitants per city.
Table 1. Number of Sentinel-5P scenes used per month and year.
Month 2019 2020
January 426 424
February 382 379
March 426 426
April 403 407
Total 1637 1636
3. Results
European countries have established differing lockdown measures in
order to control the spread of the SARS-CoV-2 outbreak. These measures,
imposed at varying degrees, were implemented at different moments
during the second half of March 2020 and have restricted freedom of
movement and outlawed public meetings. Italy and Spain were the first
countries in Europe to implement these measures on 11 and 14 March 2020,
respectively. Taking 15 March 2020 as a time reference, Figure 3 shows the
evolution of NO2 VCD in the troposphere at a European scale one month
before and after the adoption of the lockdown measures and compares it to
the concentration in the same period during the year 2019. At this scale of
detail, it can be seen that the highest NO2 VCD distribution values appears
in central Europe, large European cities and some urban areas of the
Mediterranean basin. One month before March 15, both in 2019 (Figure
3a.I) and 2020 (Figure 3b.I), the regional distribution of the average NO2
VCD around Milan, Paris, Madrid, London showed values higher than
0.0002 mol m−2. Likewise, between 15 March and 15 April 2019 (Figure
3a.II), NO2 VCD distribution was maintained around the previously
Capítulo 3
pág. 110
described urban areas. However, for the same period in 2020 (Figure 3b.II)
these extreme values greatly declined, coinciding with the lockdown
period.
Figure 3. Comparison in Europe of the evolution of the average NO2
content in the troposphere one-month (I) before and (II) after March 15,
(a) 2019 and the same period in (b) 2020.
Figure 4 shows the box and whisker plot and Table 2 the statistic
description of NO2 VCD in Spain for 2019 and 2020 from January to April.
A wide range of NO2 VCD values were observed (Figure 4a), especially in
January and February of both years. The maximum values were reduced in
March and April, especially in 2020, and declined sharply in April 2020,
coinciding with mobility restrictions. On the other hand, the minimum
values did not fluctuate, maintaining similarity throughout the months of
both years. Median, quartile 25% and 75% values were closer to the
minimum values, showing the presence of geographical areas with higher
values of NO2 VCD than the rest. Figure 4.a shows how the range of NO2
was reduced in March and April in the two years. However, while in 2019
the distribution was very similar, this was not the case in 2020. In March
2020, the reduction in the range of NO2 VCD was more pronounced than in
the same month of the previous year. In April 2020, the reduction was much
Capítulo 3
pág. 111
more pronounced, coinciding with mobility restrictions. Figure 4b shows
in detail the evolution of the distribution of NO2 VCD around the median
values. During 2019, the median values were similar, ranging from 2.08 ×
10−5 to 2.19 × 10−5 mol m−2 in January and February, respectively.
Contrariwise, in 2020, the median values of NO2 did not show the same
stable behavior of the previous year. January 2020 showed the highest
value, 2.75 × 10−5 mol m−2, reducing slightly in February, although it was
still higher than the previous year’s values. However, in the month of
March 2020 there was a very pronounced reduction of NO2 in April 2020,
with a median value equal to 1.65 × 10−5 mol m−2, the lowest value of all the
months analyzed. In addition, the interquartile range in the months of
March and April 2020 was smaller and therefore the distribution of NO2
was more homogeneous throughout the territory.
Capítulo 3
pág. 112
Figure 4. Whisker box plot with the monthly evolution of NO2 in 2019
and 2020 taking into account (a) the whole range of values and (b) zoom
on the median values.
Table 2. Descriptive statistics of NO2 concentration by month and year.
Month
Yea
r
Minimu
m
Maximu
m Q25 Median Q75
January 2019 1.46 × 10−9 0.00027
1.64 ×
10−5
2.08 ×
10−5
2.82 ×
10−5
2020 7.63 × 10−7 0.000265
2.32 ×
10−5
2.75 ×
10−5
3.37 ×
10−5
Februar
y 2019 7.41 × 10−9 0.000269
1.66 ×
10−5
2.19 ×
10−5
2.93 ×
10−5
2020 2.4 × 10−7 0.000232
1.97 ×
10−5
2.39 ×
10−5
2.98 ×
10−5
March 2019 9.56 × 10−7 0.000169
1.76 ×
10−5
2.16 ×
10−5
2.71 ×
10−5
2020 4.12 × 10−7 0.000134
1.65 ×
10−5
1.94 ×
10−5
2.29 ×
10−5
April 2019 3.32 × 10−7 0.000148 1.8 × 10−5
2.15 ×
10−5 2.6 × 10−5
2020 1.25 × 10−8 6.49E-05
1.35 ×
10−5
1.65 ×
10−5
1.99 ×
10−5
Figure 5 shows the temporal evolution of the spatial distribution of
NO2 VCD in the January–April period for the years 2019 and 2020 in Spain.
During the year 2019 (Figure 5I.x), Madrid and the surrounding
metropolitan area was the one with the highest values of NO2 VCD. In
addition, the metropolitan areas of Barcelona and Valencia, in the
Mediterranean basin, and Seville in southern Spain, stand out in NO2 VCD
values, although much lower than Madrid. These areas correspond to the
areas with the highest NO2 VCD values represented in the box and whisker
plot of Figure 4a. As shown in this figure, these urban areas appear more
highlighted in the months of January (Figure 5I.a) and February (Figure
5I.b) than in the months of March (Figure 5I.c) and April (Figure 5I.d) in
2019, although these areas presented higher values than the rest of Spain.
In 2020 (Figure 5II.x), the distribution of NO2 VCD was similar to that of
2019 in the months of January (Figure 5II.a) and February (Figure 5II.b),
with the same urban areas standing out as in 2019 due to their increase in
NO2. In March 2020, coinciding with the middle of the month in which
mobility restriction measures were adopted, a reduction in NO2 VCD was
Capítulo 3
pág. 113
observed throughout the country, with only the geographical area around
Madrid showing higher values than the rest of Spain, although lower than
in previous months. Finally, in April 2020 (Figure 5II.d), where the mobility
restrictions measures were applied throughout the month, there was a very
marked reduction in NO2 VCD throughout the totally of Spain, with
strongly homogeneous behavior and hardly any variation.
Figure 5. Monthly maps of average values of Tropospheric vertical
column of NO2 in Spain for the years (I) 2019 and (II) 2020 during the
months of (a) January, (b) February, (c) March and (d) April.
Figure 6 presents the evolution in detail of NO2 VCD over the three
most populated cities in Spain, Madrid (Figure 6a.x), Barcelona (Figure
6b.x) and Valencia (Figure 6c.x), between January and April 2020. Of the
three cities, Madrid presented the highest NO2 VCD values at the beginning
of 2020. In the month of January, both Madrid (Figure 6a.I) and Barcelona
(Figure 6b.I), due to the size of their metropolitan areas and number of
inhabitants, showed higher values in the center of these areas, reducing
radially as the distance increases from the central area. In February, there
was a reduction in NO2 in the city of Madrid (Figure 6a.II), but not in the
cities of Barcelona (Figure 6b.II) and Valencia (Figure 6c.II). These
geographical areas presented higher values than the surrounding areas. In
March, coinciding with the limitation of mobility and activity in the middle
of the month, a reduction in NO2 was observed in the three urban areas. In
the case of Madrid (Figure 6a.III), the highest NO2 values appeared around
the city and not in the metropolitan area. A similar NO2 spatial distribution
occurred in Barcelona (Figure 6b.III). On the other hand, the city of Valencia
and its metropolitan area (Figure 6c.III) present very similar NO2 values.
The effect of mobility restrictions is very evident in the month of April 2020.
Capítulo 3
pág. 114
All analyzed urban areas showed a drastic decrease in NO2 VCD, only
Madrid (Figure 6a.IV) and, to a lesser extent, Barcelona (Figure 6b.IV)
showed a very slight increase in values with respect to the surrounding
areas for the same period, making it practically impossible to identify a
pattern associated with the urban area. In the case of Valencia (Figure 6c.IV)
this difference vanished completely. Thus, after 30 days of limitations and
restrictions in mobility and activity, the values of NO2 CDV in these urban
areas were similar to those of non-urban areas.
Figure 6. Monthly maps of average values of Tropospheric vertical
column of NO2 in (a) Madrid, (b) Barcelona and (c) Valencia for 2020
during the months of (I) January, (II) February, (III) March and (IV)
April.
Figure 7 shows the variation of NO2 VCD from one month before to
one month after 15 March 2020 in Spain. The highest NO2 VCD reductions
are represented in red, while a severe reduction is represented in yellow.
On the other hand, those areas with a low reduction are represented in
green and those areas with no discernible variation are in cyan. Throughout
the territory, a decrease can be observed after the lockdown measures, with
Capítulo 3
pág. 115
some areas showing a more pronounced decrease than others. The
variation of NO2 followed the same spatial distribution as the population
density distribution presented in Figure 1. The city of Madrid, the most
densely populated city in Spain, showed a marked reduction in NO2
concentration, reaching values of −1.56 × 10−4 mol m−2. Slightly lower values
were found in areas such as Barcelona, Valencia and some coastal urban
areas with a lower population density than Madrid. On the other hand, all
these areas were surrounded by metropolitan areas where the reduction in
NO2 was much less pronounced but also important, with values around
−0.08 × 10−4 mol m−2. Similar values were found in cities such as Seville and
its metropolitan area, Valladolid and the Ebro River corridor. The rest of
the territory, with lower population densities presented a reduction
between 0.04 × 10−4 and 0.01 × 10−4 mol m−2, lower than metropolitan areas.
Therefore, the most densely populated areas with high NO2 concentrations
showed the greatest reductions compared to those areas with low density
populations. As a result, the distribution of NO2 VCD in Spain was more
homogeneous than in previous months (Figure 5II.d).
Figure 7. Mean variation of NO2 VCD between one month before and
after 15 March 2020 in Spain.
Capítulo 3
pág. 116
Figure 8 shows the average value of tropospheric VCD of NO2
reduction in relation to the number of inhabitants, resulting in nine
categories. In general, as the number of inhabitants increases, NO2
decreases. Cities with less than 50,000 inhabitants had least significant
reduction, with an average value of −2.99 × 10−5 mol m−2. On the other hand,
those cities with more than 600,000 inhabitants were those that presented
the greatest reduction, with average values of less than −1.05 × 10−4 mol m−2.
Considering the first three categories with the lowest number of
inhabitants, the factor of increase in the reduction of tropospheric VCD of
NO2 was equal to 1.28 per 50,000 inhabitants. On the other hand, it
decreased slightly among the categories of 150,000 to 600,000 inhabitants,
being equal to 1.03 per 50,000 inhabitants.
Figure 8. Number of inhabitants versus variation of tropospheric
vertical column of NO2.
The 11 cities with more than 275,000 inhabitants in Spain, are plotted
in Figure 9 which shows a negative lineal relationship between population
size and NO2 reduction with a correlation coefficient equal to 0.73 (p-value
0.00004). A negative relationship was expected between population activity
and NO2 levels, where greater activity leads to a higher level. This
component is one of the most important in urban air pollution, with the
burning of fossil fuels such as coal, oil and gas being one of the main
sources of NO2. It is estimated that about 86% of nitrogen dioxide in
European cities are caused by fossil fuels emitted from motor vehicles [65].
That means that as populations increase, NO2 also increases. With the
lockdown measures in effect there has been almost no vehicle traffic and in
turn the concentration of NO2 has been reduced.
Capítulo 3
pág. 117
Figure 9. Number of inhabitants versus variation of tropospheric vertical
column of NO2. The red line is the fitted linear function.
4. Discussion
NO2 concentrations in megacities exceed recommendations from the
World Health Organization [13]. To control the spread of coronavirus the
large majority of people have been staying at home, maintaining social
distancing practices and working remotely [66]. As expected, the direct
consequence of industries and transportation systems shutting down was
a sudden drop in air pollutant emissions. The lockdowns have provided
researchers the opportunity to set up singular experiments based on real
data and not simulations to answer the question of what would happen if
individual transport based on fuel combustion were removed and only
those linked to public service and supply were active. Under this scenario,
around the world, many cities have looked into how air quality has
improved since the lockdown measures took place. Particularly in Europe,
NO2 emissions were highly reduced over northern Italy, Spain and the
United Kingdom [67]. It is well known there is a positive association
between NO2 concentration and population size [68]. As in other studies in
other countries [69], in Spain, NO2 concentrations are located around urban
areas, being higher as population size increases, indicating anthropogenic
sources, mainly produced by vehicles. The situation caused by the SARS-
CoV-2 pandemic has led, in the case of Spain, to an 80% reduction in traffic
and a reduction of fuel sales by 83% [70].
Urban planners, engineers and policymakers should take the studies
that have capitalized on the unique conditions presented by the pandemic
Capítulo 3
pág. 118
into account to promote new strategies to reduce air pollution and
consolidate existing ones. To this end, it is necessary to consider how the
expansion of urban areas and the complex use of land, transport patterns
and socio-economic development directly factor into our living conditions.
The expansion of urban areas has led to an increase in residential areas on
the outskirts of cities, resulting in disproportionate distances between
residential housing and places of work, generating an imbalance in
transport with a high dependence on private vehicles [71]. To date, there is
not a sufficient number of studies taking into account the integration of
transport systems and urban planning to reduce air pollution [72,73].
Regardless of the positive impact on the reduction of air pollutants,
climate effects are still present today and should not be understood as a
substitute for climate change. In this way, we would like to express that in
this manuscript we characterized the changes produced on air quality
during the lockdown. We have not tried to attribute specifically nor
quantify the effects of the lockdown since other factors may have
influenced the changes, such as meteorology and regional and long-range
transport of pollutants. An in-depth analysis is required to obtain this
information accurately.
In addition to the impact of lockdown measures on air quality, future
work should aim to study the impact of the absence of tourists on the
appearance of beaches and water quality, or the reduction of noise
pollution. Moreover, the massive use of personal protective equipment
such as masks or gloves has increased worldwide, and probably recycling
and waste management policies should be analyzed and redesigned.
5. Conclusions
Lockdown measures due to SARS-CoV-2 have provoked a singular
and unique opportunity to evaluate the contribution and impact of human
activity on the environment. In this manuscript, the effect on air pollution
due to the pandemic response in Spain has been shown by evaluating and
analyzing the concentration of tropospheric NO2 from 1 January to 30 April
in 2019 and 2020. In this study, data from the TROPOMI on-board the
Sentinel-5P satellite platform were used to analyze the spatial-temporal
variation in Spain and its relationship with population size and lockdown
measures.
The satellite scenes showed a high concentration of NO2 in the city of
Madrid, which has the largest number of inhabitants in Spain. Other
hotspots with high concentrations of NO2 also appeared over cities with a
large number of inhabitants. Previous lockdown measures, the relationship
Capítulo 3
pág. 119
between the population density map of Spain and the NO2 distribution
followed the same pattern. Furthermore, as a result of the concentration of
the population in very specific points within the Spanish territory and the
source of NO2, mainly related to vehicle traffic, the distribution of this
pollutant presented a wide range of values in the air, clearly differentiating
between areas with a high population density and those without.
Just two weeks of lockdown measures and the mobility restrictions
were reflected in a reduction of NO2, mainly in those areas with larger
populations. One month later, the NO2 reduction was more evident, and
not only in the populated areas, adopting a homogeneous distribution
throughout the territory. The comparison between the NO2 concentration
values before and after the lockdown measures shows a strong relationship
with the number of inhabitants.
These results should be taken into account by governments and
policymakers to develop effective NO2 emissions reduction and air
pollution prevention policies. These policies should be based on adopting
local measures within a global project.
Author Contributions: F.-J.M.-C., A.G.-F. and J.E.M.-L. conceived and designed the
experiment, F.-J.M.-C., F.P.P. and P.T.-T. performed the experiment; F.-J.M.-C.,
F.P.P. and P.T.-T. analyzed the data and F.-J.M.-C. wrote the paper and A.G.-F. and
J.E.M.-L. collaborated in the discussion of the results and revised the manuscript.
All the authors have read and approved the manuscript.
Funding: This research received no external funding.
Conflicts of Interest: The authors declare no conflict of interest.
References
1. World Health Organization Air pollution. Available online: (accessed on).
2. Gurjar, B.R.; Lelieveld, J. New Directions: Megacities and global change.
AtmEn 2005, 39, 391–393.
3. CAI-Asia Center. Indonesia: Air Quality Profile; Clean Air Initiative for Asian
Cities (CAI-Asia) Center: Pasig, Philippines, 2010.
4. Khandelwal, S.; Goyal, R.; Kaul, N.; Mathew, A. Assessment of land surface
temperature variation due to change in elevation of area surrounding Jaipur,
India. Egypt. J. Remote Sens. Space Sci. 2018, 21, 87–94.
5. Dadhich, P.N.; Hanaoka, S. Spatial investigation of the temporal urban form
to assess impact on transit services and public transportation access. Geo Spat.
Inf. Sci. 2012, 15, 187–197.
6. Ravindra, K.; Mor, S.; Ameena; Kamyotra, J.S.; Kaushik, C.P. Variation in
Spatial Pattern of Criteria Air Pollutants Before and During Initial Rain of
Monsoon. Environ. Monit. Assess. 2003, 87, 145–153.
Capítulo 3
pág. 120
7. Marsh, W.M.; Grossa, J., Jr. Environmental Geography: Science, Land Use, and
Earth Systems, 2nd ed.; John Wiley and Sons, Ed.; John Wiley and Sons: New
York, NY, USA, 2002; ISBN 0471503967.
8. Pasqua, L.A.; Damasceno, M.V.; Cruz, R.; Matsuda, M.; Martins, M.A.G.;
Marquezini, M.V.; Lima-Silva, A.E.; Saldiva, P.H.N.; Bertuzzi, R. Exercising in
the urban center: Inflammatory and cardiovascular effects of prolonged
exercise under air pollution. Chemosphere 2020, 254, 126817.
9. Shmuel, S.; White, A.J.; Sandler, D.P. Residential exposure to vehicular traffic-
related air pollution during childhood and breast cancer risk. Environ. Res.
2017, 159, 257–263.
10. Kopnina, H. Vehicular air pollution and asthma: Implications for education
for health and environmental sustainability. Local Environ. 2017, 22, 38–48.
11. Tang, G.; Zhao, P.; Wang, Y.; Gao, W.; Cheng, M.; Xin, J.; Li, X.; Wang, Y.
Mortality and air pollution in Beijing: The long-term relationship. Atmos.
Environ. 2017, 150, 238–243.
12. He, G.; Fan, M.; Zhou, M. The effect of air pollution on mortality in China:
Evidence from the 2008 Beijing Olympic Games. J. Environ. Econ. Manag. 2016,
79, 18–39.
13. World Health Organization. WHO Air Quality Guidelines for Particulate Matter,
Ozone, Nitrogen Dioxide and Sulfur Dioxide: Global Update 2005: Summary of Risk
Assessment; World Health Organization: Geneva, Switzerland, 2006.
14. UNION, P. Directive 2008/50/EC of the European Parliament and of the
Council of 21 May 2008 on ambient air quality and cleaner air for Europe. Off.
J. Eur. Union 2008, 152, 1–44.
15. Yebin, T.; Wei, H.; Xiaoliang, H.; Liuju, Z.; Shou-En, L.; Yi, L.; Lingzhen, D.;
Yuanhang, Z.; Tong, Z. Estimated Acute Effects of Ambient Ozone and
Nitrogen Dioxide on Mortality in the Pearl River Delta of Southern China.
Environ. Health Perspect. 2012, 120, 393–398.
16. MacIntyre, E.A.; Gehring, U.; Mölter, A.; Fuertes, E.; Klümper, C.; Krämer, U.;
Quass, U.; Hoffmann, B.; Gascon, M.; Brunekreef, B.; et al. Air Pollution and
Respiratory Infections during Early Childhood: An Analysis of 10 European
Birth Cohorts within the ESCAPE Project. Environ. Health Perspect. 2014, 122,
107–113.
17. Hesterberg, T.W.; Bunn, W.B.; McClellan, R.O.; Hamade, A.K.; Long, C.M.;
Valberg, P.A. Critical review of the human data on short-term nitrogen
dioxide (NO2) exposures: Evidence for NO2 no-effect levels. Crit. Rev. Toxicol
2009, 39, 743–781.
18. Ilan, L.; Cristian, M.; Gang, L.; Julie, N.; Brook J.R. Evaluating Multipollutant
Exposure and Urban Air Quality: Pollutant Interrelationships, Neighborhood
Variability, and Nitrogen Dioxide as a Proxy Pollutant. Environ. Health
Perspect. 2014, 122, 65–72.
19. Stieb, D.M.; Burnett, R.T.; Smith-Doiron, M.; Brion, O.; Shin, H.H.; Economou,
V. A New Multipollutant, No-Threshold Air Quality Health Index Based on
Capítulo 3
pág. 121
Short-Term Associations Observed in Daily Time-Series Analyses. J. Air Waste
Manag. Assoc. 2008, 58, 435–450.
20. Filleul, L.; Rondeau, V.; Vandentorren, S.; Le Moual, N.; Cantagrel, A.; Annesi-
Maesano, I.; Charpin, D.; Declercq, C.; Neukirch, F.; Paris, C.; et al. Twenty
five year mortality and air pollution: Results from the French PAARC survey.
Occup. Environ. Med. 2005, 62, 453–460.
21. Chen, X.; Zhang, L.; Huang, J.; Song, F.; Zhang, L.; Qian, Z.; Trevathan, E.;
Mao, H.; Han, B.; Vaughn, M.; et al. Long-term exposure to urban air pollution
and lung cancer mortality: A 12-year cohort study in Northern China. Sci. Total
Environ. 2016, 571, 855–861.
22. Gauderman, W.J.; Avol, E.; Lurmann, F.; Kuenzli, N.; Gilliland, F.; Peters, J.;
McConnell, R. Childhood Asthma and Exposure to Traffic and Nitrogen
Dioxide. Epidemiology 2005, 16, 737–743.
23. Kowalska, M.; Skrzypek, M.; Kowalski, M.; Cyrys, J. Effect of NOx and NO2
Concentration Increase in Ambient Air to Daily Bronchitis and Asthma
Exacerbation, Silesian Voivodeship in Poland. Int. J. Environ. Res. Public Health
2020, 17, 754.
24. Beelen, R.; Hoek, G.; Van Den Brandt, P.A.; Goldbohm, R.A.; Fischer, P.;
Schouten, L.J.; Jerrett, M.; Hughes, E.; Armstrong, B.; Brunekreef, B. Long-
Term Effects of Traffic-Related Air Pollution on Mortality in a Dutch Cohort
(NLCS-AIR Study). Environ. Health Perspect. 2008, 116, 196–202.
25. Eum, K.-D.; Kazemiparkouhi, F.; Wang, B.; Manjourides, J.; Pun, V.; Pavlu, V.;
Suh, H. Long-term NO2 exposures and cause-specific mortality in American
older adults. Environ. Int. 2019, 124, 10–15.
26. Amorim, L.C.A.; Carneiro, J.P.; Cardeal, Z.L. An optimized method for
determination of benzene in exhaled air by gas chromatography–mass
spectrometry using solid phase microextraction as a sampling technique. J.
Chromatogr. B 2008, 865, 141–146.
27. Ma, Y.; Richards, M.; Ghanem, M.; Guo, Y.; Hassard, J. Air pollution
monitoring and mining based on sensor grid in London. Sensors 2008, 8, 3601–
3623.
28. Richards, M.; Ghanem, M.; Osmond, M.; Guo, Y.; Hassard, J. Grid-based
analysis of air pollution data. Ecol. Model. 2006, 194, 274–286.
29. Boubrima, A.; Bechkit, W.; Rivano, H. Optimal WSN Deployment Models for
Air Pollution Monitoring. IEEE Trans. Wirel. Commun. 2017, 16, 2723–2735.
30. Patil, D.; Thanuja, T.C.; Melinamath, B.C. Air Pollution Monitoring System Using
Wireless Sensor Network (WSN) BT-Data Management, Analytics and Innovation;
Balas, V.E., Sharma, N., Chakrabarti, A., Eds.; Springer: Singapore, Singapore,
2019; pp. 391–400.
31. Yi, W.Y.; Lo, K.M.; Mak, T.; Leung, K.S.; Leung, Y.; Meng, M.L. A survey of
wireless sensor network based air pollution monitoring systems. Sensors 2015,
15, 31392–31427.
Capítulo 3
pág. 122
32. Zheng, Z.; Yang, Z.; Wu, Z.; Marinello, F. Spatial Variation of NO2 and Its
Impact Factors in China: An Application of Sentinel-5P Products. Remote Sens.
2019, 11, 1939.
33. Nate, S. Remote-Sensing Applications for Environmental Health Research.
Environ. Health Perspect. 2014, 122, A268–A275.
34. Burrows, J.P.; Weber, M.; Buchwitz, M.; Rozanov, V.; Ladstätter-
Weißenmayer, A.; Richter, A.; DeBeek, R.; Hoogen, R.; Bramstedt, K.;
Eichmann, K.-U.; et al. The Global Ozone Monitoring Experiment (GOME):
Mission Concept and First Scientific Results. J. Atmos. Sci. 1999, 56, 151–175.
35. Bovensmann, H.; Burrows, J.P.; Buchwitz, M.; Frerick, J.; Noël, S.; Rozanov,
V.V.; Chance, K.V.; Goede, A.P.H. SCIAMACHY: Mission Objectives and
Measurement Modes. J. Atmos. Sci. 1999, 56, 127–150.
36. Callies, J.; Corpaccioli, E.; Eisinger, M.; Hahne, A.; Lefebvre, A. GOME-2-
Metop’s second-generation sensor for operational ozone monitoring. ESA
Bull. 2000, 102, 28–36.
37. Levelt, P.F.; Van Den Oord, G.H.J.; Dobber, M.R.; Malkki, A.; Visser, H.; De
Vries, J.; Stammes, P.; Lundell, J.O.V.; Saari, H. The ozone monitoring
instrument. IEEE Trans. Geosci. Remote Sens. 2006, 44, 1093–1101.
38. Veefkind, J.P.; Aben, I.; McMullan, K.; Förster, H.; De Vries, J.; Otter, G.; Claas,
J.; Eskes, H.J.; De Haan, J.F.; Kleipool, Q.; et al. TROPOMI on the ESA Sentinel-
5 Precursor: A GMES mission for global observations of the atmospheric
composition for climate, air quality and ozone layer applications. Remote Sens.
Environ. 2012, 120, 70–83.
39. Griffin, D.; Zhao, X.; McLinden, C.A.; Boersma, F.; Bourassa, A.; Dammers, E.;
Degenstein, D.; Eskes, H.; Fehr, L.; Fioletov, V.; et al. High-Resolution
Mapping of Nitrogen Dioxide With TROPOMI: First Results and Validation
Over the Canadian Oil Sands. Geophys. Res. Lett. 2019, 46, 1049–1060.
40. Lamsal, L.N.; Duncan, B.N.; Yoshida, Y.; Krotkov, N.A.; Pickering, K.E.;
Streets, D.G.; Lu, Z.U.S. NO2 trends (2005–2013): EPA Air Quality System
(AQS) data versus improved observations from the Ozone Monitoring
Instrument (OMI). Atmos. Environ. 2015, 110, 130–143.
41. Van Geffen, J.H.G.M.; Eskes, H.J.; Boersma, K.F.; Maasakkers, J.D.; Veefkind,
J.P. TROPOMI ATBD of the Total and Tropospheric NO2 Data Products.
Minist. Infrastruct. Water Manag. 2019. Available online: https://sentinel. esa.
int/documents/247904/2476257/Sentinel-5P-TROPOMI-ATBD-NO2-data-
products (accessed on 10 January 2020).
42. Curier, R.L.; Kranenburg, R.; Segers, A.J.S.; Timmermans, R.M.A.; Schaap, M.
Synergistic use of OMI NO2 tropospheric columns and LOTOS–EUROS to
evaluate the NOx emission trends across Europe. Remote Sens. Environ. 2014,
149, 58–69.
43. Castellanos, P.; Boersma, K.F. Reductions in nitrogen oxides over Europe
driven by environmental policy and economic recession. Sci. Rep. 2012, 2, 265.
Capítulo 3
pág. 123
44. Ghude, S.D.; Pfister, G.G.; Jena, C.; Van Der A., R.J.; Emmons, L.K.; Kumar, R.
Satellite constraints of nitrogen oxide (NOx) emissions from India based on
OMI observations and WRF-Chem simulations. Geophys. Res. Lett. 2013, 40,
423–428.
45. Streets, D.G.; Canty, T.; Carmichael, G.R.; De Foy, B.; Dickerson, R.R.; Duncan,
B.N.; Edwards, D.P.; Haynes, J.A.; Henze, D.K.; Houyoux, M.R.; et al.
Emissions estimation from satellite retrievals: A review of current capability.
Atmos. Environ. 2013, 77, 1011–1042.
46. Wang, S.W.; Zhang, Q.; Streets, D.G.; He, K.B.; Martin, R.V.; Lamsal, L.N.;
Chen, D.; Lei, Y.; Lu, Z. Growth in NOx emissions from power plants in China:
Bottom-up estimates and satellite observations. Atmos. Chem. Phys. 2012, 12,
4429.
47. Kim, S.-W.; Heckel, A.; McKeen, S.A.; Frost, G.J.; Hsie, E.-Y.; Trainer, M.K.;
Richter, A.; Burrows, J.P.; Peckham, S.E.; Grell, G.A. Satellite-observed U.S.
power plant NOx emission reductions and their impact on air quality. Geophys.
Res. Lett. 2006, 33.
48. World Economic Forum The Global Risks Report 2020. 2020. Available online:
(accessed on).
49. Ye, Z.-W.; Yuan, S.; Yuen, K.-S.; Fung, S.-Y.; Chan, C.-P.; Jin, D.-Y. Zoonotic
origins of human coronaviruses. Int. J. Biol. Sci. 2020, 16, 1686–1697.
50. Ahmad, T.; Khan, M.; Haroon; Musa, T.H.; Nasir, S.; Hui, J.; Bonilla-Aldana,
D.K.; Rodriguez-Morales, A.J. COVID-19: Zoonotic aspects. Travel Med. Infect.
Dis. 2020, doi:10.1016%2Fj.tmaid.2020.101607
51. Mackenzie, J.S.; Chua, K.B.; Daniels, P.W.; Eaton, B.T.; Field, H.E.; Hall, R.A.;
Halpin, K.; Johansen, C.A.; Kirkland, P.D.; Lam, S.K.; et al. Emerging viral
diseases of Southeast Asia and the Western Pacific. Emerg. Infect. Dis. 2001, 7,
497–504.
52. Olsen, B.; Munster, V.J.; Wallensten, A.; Waldenström, J.; Osterhaus, A.D.M.E.;
Fouchier, R.A.M. Global Patterns of Influenza a Virus in Wild Birds. Science
2006, 312, 384–388.
53. Fergus, R.; Fry, M.; Karesh, W.B.; Marra, P.P.; Newman, S.; Paul, E. Migratory
Birds and Avian Flu. Science 2006, 312, 845–846.
54. Petersen, L.R.; Marfin, A.A. Shifting Epidemiology of Flaviviridae. J. Travel
Med. 2008, 12, s3–s11.
55. Leviston, Z.; Leitch, A.; Greenhill, M.; Leonard, R.; Walker, I. Australians’
views of climate change. Canberra CSIRO 2011.
56. Price, J.C.; Walker, I.A.; Boschetti, F. Measuring cultural values and beliefs
about environment to identify their role in climate change responses. J.
Environ. Psychol. 2014, 37, 8–20.
57. Van Der Linden, S.L.; Leiserowitz, A.A.; Feinberg, G.D.; Maibach, E.W. The
Scientific Consensus on Climate Change as a Gateway Belief: Experimental
Evidence. PLoS ONE 2015, 10, e0118489.
Capítulo 3
pág. 124
58. Callaway, E. Time to use the p-word? Coronavirus enter dangerous new
phase. Nature 2020, 579, 10–38.
59. Remuzzi, A.; Remuzzi, G. COVID-19 and Italy: What next? Lancet 2020, 395,
1225–1228.
60. Hou, C.; Chen, J.; Zhou, Y.; Hua, L.; Yuan, J.; He, S.; Guo, Y.; Zhang, S.; Jia, Q.;
Zhao, C.; et al. The effectiveness of quarantine of Wuhan city against the
Corona Virus Disease 2019 (COVID-19): A well-mixed SEIR model analysis. J.
Med. Virol. 2020, 92, 841–848.
61. Lau, H.; Khosrawipour, V.; Kocbach, P.; Mikolajczyk, A.; Schubert, J.; Bania,
J.; Khosrawipour, T. The positive impact of lockdown in Wuhan on containing
the COVID-19 outbreak in China. J. Travel Med. 2020, 27,
doi:10.1093/jtm/taaa037.
62. Peto, J.; Alwan, N.A.; Godfrey, K.M.; Burgess, R.A.; Hunter, D.J.; Riboli, E.;
Romer, P. Universal weekly testing as the UK COVID-19 lockdown exit
strategy. Lancet 2020, 395, 1420–1421.
63. Gorelick, N.; Hancher, M.; Dixon, M.; Ilyushchenko, S.; Thau, D.; Moore, R.
Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote
Sens. Environ. 2017, 202, 18–27.
64. Eskes, H.J.; Eichmann, K.U. S5P Mission Performance Centre Nitrogen Dioxide
[L2__NO2___]. 2019. Available online: (accessed on).
65. Degraeuwe, B.; Pisoni, E.; Peduzzi, E.; De Meij, A.; Monforti-Ferrario, F.;
Bodis, K.; Mascherpa, A.; Astorga-Llorens, M.; Thunis, P.; Vignati, E. Urban
NO2 Atlas; Publications Office of the European Union: Brussels, Belgium, 2019;
ISBN 978-92-76-10386-8.
66. Harapan, H.; Itoh, N.; Yufika, A.; Winardi, W.; Keam, S.; Te, H.; Megawati, D.;
Hayati, Z.; Wagner, A.L.; Mudatsir, M. Coronavirus disease 2019 (COVID-19):
A literature review. J. Infect. Public Health 2020, 13, 667–673.
67. Ficetola, G.F.; Rubolini, D. Climate affects global patterns of COVID-19 early
outbreak dynamics. MedRxiv 2020, doi:10.1101/2020.03.23.20040501.
68. Zhu, Y.; Price, O.R.; Kilgallon, J.; Qi, Y.; Tao, S.; Jones, K.C.; Sweetman, A.J.
Drivers of contaminant levels in surface water of China during 2000–2030:
Relative importance for illustrative home and personal care product
chemicals. Environ. Int. 2018, 115, 161–169.
69. Zhu, Y.; Zhan, Y.; Wang, B.; Li, Z.; Qin, Y.; Zhang, K. Spatiotemporally
mapping of the relationship between NO2 pollution and urbanization for a
megacity in Southwest China during 2005–2016. Chemosphere 2019, 220, 155–
162.
70. De Tráfico, D.G.; Del, I.M. Evolución del Tráfico por el efecto COVID-19.
Available online: http://www.dgt.es/Galerias/covid-19/Evolucion-
Intensidades-dia-02-04-2020-Periodo-Coronavirus.pdf (accessed on 9 May
2020).
71. Banister, D. Energy, quality of life and the environment: The role of transport.
Transp. Rev. 1996, 16, 23–35.
Capítulo 3
pág. 125
72. Camagni, R.; Gibelli, M.C.; Rigamonti, P. Urban mobility and urban form: The
social and environmental costs of different patterns of urban expansion. Ecol.
Econ. 2002, 40, 199–216.
73. Ambarwati, L.; Verhaeghe, R.; Van Arem, B.; Pel, A.J. The influence of
integrated space–transport development strategies on air pollution in urban
areas. Transp. Res. Part D Transp. Environ. 2016, 44, 134–146.
© 2020 by the authors. Submitted for possible open access
publication under the terms and conditions of the Creative
Commons Attribution (CC BY) license
(http://creativecommons.org/licenses/by/4.0/).
Conclusiones
Las características de los vehículos aéreos no tripulados por su baja
altitud de vuelo, su bajo coste y gran flexibilidad, ofrecen nuevas
oportunidades para las aplicaciones de agricultura de precisión
aumentando la resolución espacial y temporal con datos de múltiples
fuentes. Aun así, uno de los puntos débiles de los UAV comerciales y de
bajo coste es hasta ahora la autonomía de vuelo, lo que implica la
miniaturización máxima de la electrónica de las cargas de pago, para
aumentar el rendimiento de la autonomía lo máximo posible.
A pesar de que las plataformas no tripuladas con sensores son
fácilmente accesibles en la actualidad y se han usado ampliamente en una
gran cantidad de proyectos con éxitos diversos, todavía es necesario seguir
trabajando en el pre-procesamiento de los datos capturados por las
distintas fuentes con objeto de poder realizar análisis concluyentes sobre
los mismos.
Por otro lado, la reducción en tamaño de los componentes electrónicos
genera problemas que son necesarios resolver con metodologías
estandarizadas para su posterior aplicación sobre los cultivos agrarios.
Cada tipo de sensor pasivo miniaturizado necesita de correcciones para
ofrecer resultados similares a los sensores embarcados en plataformas
aéreas tripuladas o satelitales. Las técnicas tradicionales de pre-procesado
y corrección de captura de datos se están actualizando por procedimientos
más avanzados a partir de modelos matemáticos que explican y corrigen
su comportamiento.
En general, como se ha comprobado con los resultados obtenidos en el
capítulo 1 de la tesis, una nueva metodología de pre-procesamiento de
datos ha sido aplicada a los datos termográficos capturados por el sensor
embarcado en el UAV. Una vez que los datos han sido pre-procesados se
han aplicado metodologías clásicas para las correcciones atmosféricas, que
en este caso si han surtido efecto. En definitiva, los datos capturados por
plataformas UAV menores de 25 kg con pequeños sensores necesitan de un
pre-procesado para posteriormente aplicar sobre ellos técnicas clásicas de
teledetección.
Cada vez es más importante divulgar este tipo de pre-procesado de
datos en todos los ámbitos, académico, empresarial, industrial, etc. Debido
a la velocidad a la que avanza la tecnología, es necesario que se transfiera
Conclusiones
pág. 128
de forma rápida este conocimiento como se ha visto en el capítulo 2, para
que las técnicas aplicadas sobre los datos capturados de UAV sean las
correctas e incorporen esta fase en el procesamiento.
Una vez que se han corregido los datos capturados por estos sensores
de UAV a partir de este nuevo concepto para la modelización de la deriva
térmica, el flujo de trabajo para generar productos a través de teledetección
sigue una estructura fija y estandarizada que permite a partir de datos
capturados por un sensor y datos terreno, generar modelos predictivos.
Finalmente, en el capítulo 3 de la tesis, se ha seguido un flujo de trabajo
para la predicción de variables ambientales desde satélite de forma análoga
a como se realiza a partir de datos capturados por sensores de UAV, una
vez pre-procesados. Esto es de gran importancia ya que identifica un flujo
de trabajo estándar para teledetección donde la única diferencia entre el
procesado de datos desde UAV y satélite, sólo implica un pre-procesado
inicial en los datos capturados desde plataformas no tripuladas.